Suitability of a Commercial Software Defined Radio System for ...
Suitability of a Commercial Software Defined Radio System for ...
Suitability of a Commercial Software Defined Radio System for ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>Suitability</strong> <strong>of</strong> a <strong>Commercial</strong> S<strong>of</strong>tware <strong>Defined</strong> <strong>Radio</strong><br />
<strong>System</strong> <strong>for</strong> Passive Coherent Location<br />
Aadil Volkwin<br />
A dissertation submitted to the Department <strong>of</strong> Electrical Engineering,<br />
University <strong>of</strong> Cape Town,<br />
in fulfilment <strong>of</strong> the requirements <strong>for</strong> the degree <strong>of</strong><br />
Master <strong>of</strong> Science in Engineering.<br />
Cape Town, May 2008
Dedicated to:<br />
My Parents,<br />
my Sister and my Wife.<br />
2
Declaration<br />
I declare that this work done is my own, unaided work. This dissertation is being submitted to<br />
the Department <strong>of</strong> Electrical Engineering, University <strong>of</strong> Cape Town, in fulfilment <strong>of</strong> the<br />
requirements <strong>for</strong> the degree <strong>of</strong> Master <strong>of</strong> Science in Engineering. It has not been submitted <strong>for</strong><br />
any degree or examination in any other university.<br />
................................................................................<br />
Signature <strong>of</strong> Author<br />
Cape Town<br />
2008<br />
3
Abstract<br />
This dissertation provides a comprehensive discussion around bistatic radar with specific<br />
reference to PCL, highlighting existing literature and work, examining the various per<strong>for</strong>mance<br />
metrics. In particular the per<strong>for</strong>mance <strong>of</strong> commercial FM radio broadcasts as the radar wave<strong>for</strong>m<br />
is examined by implementation <strong>of</strong> the ambiguity function.<br />
The FM signals show desirable characteristics in the context <strong>of</strong> our application, the average<br />
range resolution obtained is 5.98km, with range and doppler peak sidelobe levels measured at<br />
-25.98dB and -33.14dB respectively.<br />
Furthermore, the SDR paradigm and technology is examined, with discussion around the design<br />
considerations. The USRP, the TVRx daughterboard and GNU<strong>Radio</strong> are examined further as a<br />
potential receiver and development environment, in this light.<br />
The system meets the low cost ambitions costing just over US$1000.00 <strong>for</strong> the USRP<br />
motherboard and a single daughterboard. Furthermore it per<strong>for</strong>ms well, displaying desirable<br />
characteristics, The receiver's frontend provides a bandwidth <strong>of</strong> 6MHz and a tunable range<br />
between 50MHz and 800MHz, with a tuning step size as low as 31.25kHz. The noise<br />
characterisation <strong>of</strong> the receiver reveals a NF <strong>of</strong> 10dB, a sensitivity <strong>of</strong> -105dB and a dynamic<br />
range <strong>of</strong> 62dB.<br />
Finally, the investigation into the stability <strong>of</strong> the daughterboard frontend oscillators due to ageing<br />
effects is shown to be steady with acceptable levels <strong>of</strong> variation, showing a fractional frequency<br />
variation from 2.3 to 0.8 parts per 10 million and a maximum frequency drift <strong>of</strong> 4Hz.<br />
4
Acknowledgements<br />
All praises and thanks to the almighty <strong>for</strong> all the blessings and abilities that have been af<strong>for</strong>ded<br />
to me.<br />
Further gratitude is expressed to:<br />
My supervisor, Pr<strong>of</strong>. M.R. Inggs <strong>for</strong> his advice, support and encouragement.<br />
Dr. Yoann Paichard, <strong>for</strong> invaluable technical advice in developing and interpreting various<br />
aspects <strong>of</strong> the research.<br />
To the RRSG, in particular Regine Lord <strong>for</strong> administrative support, Lance Williams, Jonathan<br />
Ward and Gunther Lange <strong>for</strong> various hardware assistance.<br />
To members <strong>of</strong> the GNU<strong>Radio</strong> mailing list, in particular Matt Ettus and Firas Abbas.<br />
Finally, to my family and friends <strong>for</strong> their encouragement, support and patience .<br />
5
Contents<br />
Declaration.......................................................................................................................................3<br />
Abstract............................................................................................................................................4<br />
Acknowledgements..........................................................................................................................5<br />
Contents........................................................................................................................................... 6<br />
List <strong>of</strong> Figures..................................................................................................................................8<br />
List <strong>of</strong> Tables................................................................................................................................. 10<br />
Nomenclature.................................................................................................................................11<br />
Chapter 1........................................................................................................................................13<br />
Introduction................................................................................................................................... 13<br />
1.1 Background........................................................................................................................ 13<br />
1.2 Plan <strong>of</strong> Development..........................................................................................................14<br />
Chapter 2........................................................................................................................................20<br />
Passive Coherent Location............................................................................................................ 20<br />
2.1 <strong>System</strong> level description <strong>of</strong> PCL........................................................................................20<br />
2.2 Advantages and Disadvantages <strong>of</strong> PCL <strong>System</strong>s...............................................................25<br />
2.3 History and Examples <strong>of</strong> PCL <strong>System</strong>s............................................................................. 26<br />
2.4 Applications <strong>of</strong> PCL.......................................................................................................... 28<br />
2.5 Typical Illuminators...........................................................................................................28<br />
2.6 Relevant Equations .......................................................................................................... 30<br />
2.7 Summary ........................................................................................................................... 40<br />
Chapter 3........................................................................................................................................41<br />
USRP and GNU<strong>Radio</strong>................................................................................................................... 41<br />
3.1 Introduction ......................................................................................................................41<br />
3.2 S<strong>of</strong>tware <strong>Defined</strong> <strong>Radio</strong> ....................................................................................................41<br />
3.3 Anatomy <strong>of</strong> a SDR Receiver..............................................................................................43<br />
3.4 Universal S<strong>of</strong>tware <strong>Radio</strong> Peripheral.................................................................................44<br />
3.5 GNU<strong>Radio</strong>......................................................................................................................... 52<br />
3.6 Summary............................................................................................................................ 55<br />
Chapter 4........................................................................................................................................57<br />
Receiver Characterisation..............................................................................................................57<br />
4.1 Introduction ......................................................................................................................57<br />
4.2 Test <strong>System</strong> Overview ...................................................................................................... 57<br />
4.3 Receiver Bandwidth...........................................................................................................58<br />
4.4 Gain Range and Control.....................................................................................................59<br />
4.5 Multiplexer Usage and Data Interleaving.......................................................................... 61<br />
4.6 Frequency Range and Control........................................................................................... 62<br />
4.7 Spurious Artifact ...............................................................................................................63<br />
4.8 Receiver Noise Figure .......................................................................................................64<br />
4.9 Sensitivity.......................................................................................................................... 66<br />
4.10 Dynamic Range, Minimum Detectable Signal and Gain Compression...........................67<br />
4.11 Summary.......................................................................................................................... 71<br />
6
Chapter 5........................................................................................................................................73<br />
Ambiguity Plot Analysis............................................................................................................... 73<br />
5.1 Introduction ......................................................................................................................73<br />
5.2 <strong>System</strong> Overview............................................................................................................... 73<br />
5.3 Algorithm Concept.............................................................................................................75<br />
5.4 Results................................................................................................................................76<br />
5.5 Summary............................................................................................................................ 82<br />
Chapter 6........................................................................................................................................84<br />
Oscillator Stability.........................................................................................................................84<br />
6.1 Introduction ......................................................................................................................84<br />
6.2 Quartz Crystal Oscillators ................................................................................................85<br />
6.3 Characterisation <strong>of</strong> stability .............................................................................................. 86<br />
6.4 <strong>System</strong> Overview .............................................................................................................. 92<br />
6.5 Results ..............................................................................................................................94<br />
6.6 Summary............................................................................................................................ 95<br />
Chapter 7........................................................................................................................................96<br />
Conclusions and Recommendations..............................................................................................96<br />
Bibliography ................................................................................................................................ 98<br />
Appendix A ................................................................................................................................100<br />
Ambiguity Analysis and Results................................................................................................. 100<br />
Appendix B ................................................................................................................................116<br />
Datasheet and Sentech Table.......................................................................................................116<br />
7
List <strong>of</strong> Figures<br />
Figure 1.1: PCL block diagram..................................................................................................... 15<br />
Figure 1.2: S<strong>of</strong>tware <strong>Radio</strong> block diagram....................................................................................15<br />
Figure 1.3: Experimental setup <strong>of</strong> the USRP <strong>for</strong> the characterisation tests...................................16<br />
Figure 1.4: USRP with two TVRx boards, reference channel and target channel........................ 18<br />
Figure 2.1: PCL system diagram................................................................................................... 22<br />
Figure 2.2: Variation <strong>of</strong> RCS and angular width <strong>of</strong> a <strong>for</strong>ward scatter target against frequency... 32<br />
Figure 2.3: A target echo detected after the transmission <strong>of</strong> the second pulse results in ambiguous<br />
ranges at B and C...........................................................................................................................36<br />
Figure 3.1: USRP and daughterboard block diagram....................................................................45<br />
Figure 3.2: USRP and daughterboards.......................................................................................... 46<br />
Figure 3.3: AD9862 ADC block Diagram.....................................................................................47<br />
Figure 3.4: USRP and FPGA block Diagram................................................................................48<br />
Figure 3.5: DDC block Diagram................................................................................................... 49<br />
Figure 3.6: USRP transmit chain block Diagram.......................................................................... 50<br />
Figure 3.7: Signal flow from a single TVRx daughterboard to DDC0......................................... 51<br />
Figure 3.8: GNU<strong>Radio</strong> application structure.................................................................................53<br />
Figure 4.1: Experimental setup used in characterisation tests.......................................................58<br />
Figure 4.2: Block diagram showing TVRx gain stages.................................................................59<br />
Figure 4.3: Variation <strong>of</strong> voltage applied to RF AGC against RF Gain......................................... 60<br />
Figure 4.4: Variation <strong>of</strong> voltage applied to IF AGC against IF Gain............................................60<br />
Figure 4.5: multiplexer setup <strong>for</strong> two TVRx to DDC0 and DDC1............................................... 61<br />
Figure 4.6: Spurious signal generated by USRP on startup...........................................................63<br />
Figure 4.7: closer look at the spurious signal generated by USRP on startup...............................64<br />
Figure 4.8: Cascaded components................................................................................................. 65<br />
Figure 4.9: Variation <strong>of</strong> receiver sensitivity..................................................................................67<br />
Figure 4.10: MDS, 1 dB compression point and dynamic range...................................................68<br />
Figure 4.11: Spurious Free Dynamic Range..................................................................................69<br />
Figure 4.12: Input power against output power, showing receiver dynamic range.......................70<br />
Figure 4.13: TVRx and ADC dynamic range comparison............................................................ 71<br />
Figure 5.1: Experiment setup to capture signals <strong>for</strong> ambiguity function analysis.........................74<br />
Figure 5.2: Block Diagram <strong>of</strong> experimental setup.........................................................................74<br />
Figure 5.3: Signal processing algorithm........................................................................................75<br />
Figure 5.4: Ambiguity plot <strong>of</strong> Rad5 transmission......................................................................... 77<br />
Figure 5.4a: Zero doppler cut through ambiguity plot.................................................................. 77<br />
Figure 5.4b: Zero delay cut through ambiguity plot......................................................................78<br />
Figure 5.5a: Variation <strong>of</strong> signal range resolution with time..........................................................81<br />
Figure 5.5b: Variation <strong>of</strong> signal range resolution with time..........................................................82<br />
Figure 6.1: Determining the fractional frequency......................................................................... 88<br />
Figure 6.2: Dependence <strong>of</strong> stability measurement on time window............................................. 89<br />
8
Figure 6.3: measurement process.................................................................................................. 91<br />
Figure 6.4: Block diagram <strong>of</strong> experimental setup......................................................................... 92<br />
Figure 6.5: Experiment setup <strong>for</strong> stability tests.............................................................................93<br />
Figure 6.6: Fractional frequency variation <strong>of</strong> the daughterboards................................................ 94<br />
Figure 6.7: Allan Variance <strong>of</strong> the fractional frequencies.............................................................. 95<br />
Figure A.1: Ambiguity plot <strong>of</strong> <strong>Radio</strong> 2000 transmission............................................................ 100<br />
Figure A.1 a: Zero doppler cut through ambiguity plot...............................................................101<br />
Figure A.1 b: Zero delay cut through ambiguity plot..................................................................101<br />
Figure A.2: Ambiguity plot <strong>of</strong> Classic FM transmission............................................................ 103<br />
Figure A.2a: Zero doppler cut through ambiguity plot................................................................103<br />
Figure A.2b: Zero delay cut through ambiguity plot...................................................................104<br />
Figure A.3: Ambiguity plot <strong>of</strong> UmhloboFM transmission..........................................................105<br />
Figure A.3a: Zero doppler cut through ambiguity plot................................................................106<br />
Figure A.3b: Zero delay cut through ambiguity plot...................................................................106<br />
Figure A.4: Ambiguity plot <strong>for</strong> RGHP transmission...................................................................108<br />
Figure A.4a: Zero doppler cut through ambiguity plot................................................................108<br />
Figure A.4b: Zero delay cut through ambiguity plot...................................................................109<br />
Figure A.5: Ambiguity plot <strong>of</strong> RSGR transmission.................................................................... 110<br />
Figure A.5a: Zero doppler cut through ambiguity plot................................................................111<br />
Figure A.5b: Zero delay cut through ambiguity plot...................................................................111<br />
Figure A.6: Ambiguity plot <strong>of</strong> SAFM transmission....................................................................113<br />
Figure A.6a: Zero doppler cut through ambiguity plot................................................................114<br />
Figure A.6b: Zero delay cut through ambiguity plot...................................................................114<br />
9
List <strong>of</strong> Tables<br />
Table 5.1: Bandwidth Variation <strong>of</strong> RAD5 wave<strong>for</strong>m................................................................... 78<br />
Table 5.2: Summary <strong>of</strong> wave<strong>for</strong>m ambiguity function per<strong>for</strong>mances...........................................83<br />
Table A.1: Bandwidth variation <strong>of</strong> <strong>Radio</strong> 2000 wave<strong>for</strong>m......................................................... 102<br />
Table A.2: Bandwidth variation <strong>of</strong> Classic FM wave<strong>for</strong>m..........................................................104<br />
Table A.3: Bandwidth variation <strong>of</strong> Umhlobo FM wave<strong>for</strong>m......................................................107<br />
Table A.4: Bandwidth variation <strong>of</strong> RGHP wave<strong>for</strong>m................................................................. 109<br />
Table A.5: Bandwidth variation <strong>of</strong> RSFR wave<strong>for</strong>m.................................................................. 112<br />
Table A.6: Bandwidth variation <strong>of</strong> SAFM wave<strong>for</strong>m.................................................................115<br />
10
Nomenclature<br />
AGC<br />
ADBF<br />
CFAR<br />
CIC<br />
CW<br />
DBS<br />
DDC<br />
DOA<br />
DPI<br />
DSP<br />
FPGA<br />
FSF<br />
GNU<br />
GPIF<br />
GPS<br />
GSM<br />
IP3<br />
LO<br />
MDS<br />
MTI<br />
NCO<br />
PCL<br />
PGA<br />
PPM<br />
RCS<br />
SAR<br />
SDR<br />
SFDR<br />
Automatic Gain Control<br />
Adaptive Beam<strong>for</strong>mer<br />
Constant False Alarm Rate<br />
Cascaded Integrator Comb Filter<br />
Continuous Wave<br />
Direct Broadcast by Satellite<br />
Digital Down Converter<br />
Direction Of Arrival<br />
Direct Path Interference<br />
Digital Signal Processors<br />
Field Programmable Gate Arrays<br />
Free S<strong>of</strong>tware Foundation<br />
GNU Not Unix<br />
General Purpose Interface<br />
Global Positioning <strong>System</strong><br />
Global <strong>System</strong> <strong>for</strong> Mobile<br />
Third Order Intercept Point<br />
Local Oscillator<br />
Minimum Detectable Signal<br />
Moving Target Indicator<br />
Numerically Controlled Oscillator<br />
Passive Coherent Location radar<br />
Programmable Gain Amplifier<br />
Part Per Million<br />
Radar Cross Section<br />
Synthetic Aperture Radar<br />
S<strong>of</strong>tware <strong>Defined</strong> <strong>Radio</strong><br />
Spurious Free Dynamic Range<br />
11
SINAD<br />
SIR<br />
SNR<br />
USRP<br />
Signal to Noise and Distortion ratio<br />
Signal to Interference Ratio<br />
Signal to Noise Ratio<br />
Universal S<strong>of</strong>tware <strong>Radio</strong> Peripheral<br />
12
Chapter 1<br />
Introduction<br />
The University <strong>of</strong> Cape Town (UCT) Radar Remote Sensing Group (RRSG) is currently<br />
engaged in the research and development <strong>of</strong> an air traffic control system that utilises passive<br />
coherent location (PCL) radar.<br />
Thus the purpose <strong>of</strong> this dissertation is to investigate the per<strong>for</strong>mance <strong>of</strong> the Universal S<strong>of</strong>tware<br />
<strong>Radio</strong> Peripheral (USRP) and the GNU Not Unix (GNU) <strong>Radio</strong> s<strong>of</strong>tware radio development<br />
environment as a low cost PCL receiver and additionally to investigate and comment on the<br />
per<strong>for</strong>mance <strong>of</strong> analogue FM radio broadcast signals as the radar wave<strong>for</strong>m.<br />
1.1 Background<br />
The first radar systems built were bistatic in nature, with transmitter and receiver in separate<br />
locations. An example <strong>of</strong> this is the Klein Heidelberg, developed by the Germans during the<br />
second world war, that used the British Chain Home radars as illuminators. Technological<br />
developments in the <strong>for</strong>m <strong>of</strong> the duplexer in 1936 allowed the transmitter and receiver to share a<br />
common antenna and there<strong>for</strong>e the same site [34].<br />
This provided the means <strong>for</strong> simplicity <strong>of</strong> operation as well as savings on cost and space and<br />
hence monostatic radar dominated radar research and design.<br />
Bistatic radar does have advantages over monostatic radar, however, their disadvantages; in<br />
particular the complexity <strong>of</strong> transmitter/receiver synchronisation and antenna beam pointing had<br />
previously outweighed their advantages and hampered the advancement <strong>of</strong> bistatic radar.<br />
Technological improvements in the shape <strong>of</strong> high speed digital signal processors (DSP), phased<br />
array antennas and the deployment <strong>of</strong> global positioning system (GPS) satellite navigation<br />
systems, which can be used <strong>for</strong> synchronisation, have provided the means to mitigate the<br />
complexities <strong>of</strong> bistatic radar and thus fuelling the current wave <strong>of</strong> interest in bistatic radar<br />
systems.<br />
Interest in bistatic radar varies on a period <strong>of</strong> fifteen years, currently we are at a peak <strong>of</strong> that<br />
cycle; with particular interest in PCL techniques [35], using 'illuminators <strong>of</strong> opportunity' such as<br />
existing radar transmissions and broadcast and communication signals.<br />
13
Of the illuminators <strong>of</strong> opportunity available in the environment, broadcast transmitters present<br />
themselves as the most attractive <strong>for</strong> surveillance purposes, owing to their inherent properties <strong>of</strong><br />
high powers <strong>of</strong> transmission and widespread coverage.<br />
If such signals are to be used on the basis <strong>of</strong> PCL, it is necessary to know their behaviour in<br />
terms <strong>of</strong> their ambiguity function [36] i.e. resolution and sidelobe levels in range and doppler,<br />
and the effect <strong>of</strong> range and doppler ambiguities.<br />
Over the last ten years the enhancement <strong>of</strong> semiconductor technologies in per<strong>for</strong>mance<br />
capability and cost has led to the emergence <strong>of</strong> S<strong>of</strong>tware <strong>Defined</strong> <strong>Radio</strong> (SDR) as a mainstream<br />
technology. SDR is defined as a collection <strong>of</strong> hardware and s<strong>of</strong>tware technologies that enable<br />
reconfigurable system architectures <strong>for</strong> wireless networks and user terminals. SDR provides an<br />
efficient and comparatively inexpensive solution to the problem <strong>of</strong> building multi-mode, multiband,<br />
multi-functional wireless devices that can be adapted, upgraded or enhanced by using<br />
s<strong>of</strong>tware upgrades [31].<br />
The obvious flexibility and costs benefits <strong>of</strong>fered by SDR systems provide the impetus <strong>for</strong> the<br />
choice to explore the USRP and GNU<strong>Radio</strong> as a radar receiver in a bistatic PCL context.<br />
1.2 Plan <strong>of</strong> Development<br />
Chapter 1:<br />
This chapter, which provides a brief background to the subject <strong>of</strong> the dissertation and motivation<br />
<strong>for</strong> the research pursued.<br />
Chapter 2:<br />
This chapter reviews the published material on PCL radar techniques to provide the background<br />
in<strong>for</strong>mation to the techniques used within the research. This chapter begins by defining PCL<br />
radar and its uses. The discussion then evolves describing the various advantages and<br />
disadvantages <strong>for</strong> the system. Thereafter, the history <strong>of</strong> bistatic radars is discussed, ranging from<br />
the first use <strong>of</strong> these systems to modern day systems. The discussion progresses further to<br />
highlight previous work in the development <strong>of</strong> a PCL radar system and examines the<br />
characteristics that motivate the choice <strong>of</strong> commercial FM radio broadcasts as a potential radar<br />
wave<strong>for</strong>m. Various applications <strong>of</strong> PCL radars were described, both <strong>for</strong> military uses as well as<br />
non military uses. These applications were presented to show how bistatic radars have been <strong>of</strong><br />
use to society. Some <strong>of</strong> the enhanced techniques <strong>of</strong> bistatic and PCL systems are also discussed<br />
with regards to the per<strong>for</strong>mance <strong>of</strong> these systems. The literature reviewed provides the necessary<br />
background understanding and sets up the context in which this research finds purpose.<br />
The following image depicts the geometrical setup <strong>of</strong> a PCL system.<br />
14
Figure 1.1: PCL block diagram<br />
Chapter 3:<br />
S<strong>of</strong>tware radio is a system which turns radio hardware problems into s<strong>of</strong>tware problems. The<br />
fundamental characteristic <strong>of</strong> s<strong>of</strong>tware radio is that s<strong>of</strong>tware defines the transmitted wave<strong>for</strong>ms<br />
and s<strong>of</strong>tware demodulates the received wave<strong>for</strong>ms. Thus a s<strong>of</strong>tware radio can tune to any<br />
frequency band and receive any modulation across a large frequency spectrum. A block diagram<br />
<strong>of</strong> the basic SDR structure is shown below.<br />
Figure 1.2: S<strong>of</strong>tware <strong>Radio</strong> block diagram<br />
15
Additionally, per<strong>for</strong>ming a significant amount <strong>of</strong> the signal processing on reprogrammable<br />
hardware permits the radios to receive and transmit a new <strong>for</strong>m <strong>of</strong> radio protocol simply by<br />
running an alternative configuration <strong>of</strong> the hardware. This is unlike most radios in which the<br />
processing is done with either analogue circuitry alone or combined with digital chips.<br />
The chapter explores the USRP and GNU<strong>Radio</strong> as development tools <strong>of</strong> this technology, in the<br />
light <strong>of</strong> this. Through this it is found that the USRP provides a flexible and powerful FPGA, the<br />
Altera cyclone, as the processing unit, which implements a great deal <strong>of</strong> the signal processing in<br />
the receive and transmit chains. Furthermore, the USRP hosts two Analog Devices AD9862<br />
chips, each providing a two ADCs sampling at a rate <strong>of</strong> 64 Msamples/second with 12 bits <strong>of</strong><br />
resolution and two DACs operating at 128 Msamples/second with 14 bits <strong>of</strong> resolution.<br />
Moreover, the AD9862 chips provide two auxiliary ADCs and three auxiliary DACs which can<br />
provide further flexibility in computation and control <strong>of</strong> external daughterboard components.<br />
These daughterboards provide front end flexibility allowing <strong>for</strong> the down-conversion and thus<br />
manipulation <strong>of</strong> the radio spectrum.<br />
GNU<strong>Radio</strong> is seen to be a powerful toolkit that implements numerous complex signal processing<br />
elements and additionally interfaces with the USRP. Thus, GNU<strong>Radio</strong> and the USRP combine to<br />
provide a powerful and widely versatile toolset <strong>for</strong> the development <strong>of</strong> a variety <strong>of</strong> radio devices.<br />
Chapter 4:<br />
Often the most critical component <strong>of</strong> a wireless system is its receiver, the purpose <strong>of</strong> which is to<br />
extract, reliably, the desired signal from the various sources <strong>of</strong> signals, interference and noise.<br />
This chapter presents the results <strong>of</strong> experiments that examine and characterise the per<strong>for</strong>mance<br />
<strong>of</strong> the USRP, TVRx daughterboard, which is a complete VHF and UHF receiver system based<br />
on a TV tuner module and GNU<strong>Radio</strong> as a PCL receiver.<br />
Figure 1.3: Experimental setup <strong>of</strong> the USRP <strong>for</strong> the characterisation tests<br />
A number <strong>of</strong> the TVRx and USRP receiver system parameters as well as the GNU<strong>Radio</strong><br />
interface to the controllable aspects are examined. The receiver's frontend provides a bandwidth<br />
<strong>of</strong> 6MHz, which far exceeds the per<strong>for</strong>mance requirements <strong>of</strong> our application. and a tunable<br />
range between 50MHz and 800MHz, with a tuning step size as low as 31.25kHz. This is<br />
sufficient <strong>for</strong> the manipulation <strong>of</strong> FM Broadcast signals as the radar wave<strong>for</strong>m and provides<br />
further functionality and versatility to incorporate additional areas <strong>of</strong> the frequency spectrum if<br />
desired, i.e. the television video and sound carriers.<br />
16
The multiplexer implemented in the FPGA and the ability to directly control its operation<br />
through GNU<strong>Radio</strong> further extends the receiver's flexibility in its digital manipulation <strong>of</strong> signals<br />
through it. The noise characterisation <strong>of</strong> the receiver reveals a NF <strong>of</strong> 10dB, a sensitivity <strong>of</strong><br />
-105dB and a dynamic range <strong>of</strong> 62dB. Furthermore, it is observed that when the GNU<strong>Radio</strong><br />
application is started, when the USRP is setup as a source and includes the occasion when the<br />
application is stopped and restarted, a spurious spike occurs at the beginning <strong>of</strong> the data set. This<br />
phenomenon is unexplained but noted by the USRP developers. The effects <strong>of</strong> this are mitigated<br />
by simply cutting out that part <strong>of</strong> the data set.<br />
Thus the USRP with the TVRx and GNU<strong>Radio</strong> show per<strong>for</strong>mance characteristics suitable <strong>for</strong><br />
our application.<br />
Chapter 5:<br />
This chapter presents the measurements <strong>of</strong> the ambiguity functions <strong>of</strong> <strong>of</strong>f-air signals that will be<br />
considered <strong>for</strong> the PCL system and comments on their <strong>for</strong>m and usefulness by demonstrating the<br />
wave<strong>for</strong>m variability and its effect on range and doppler resolution in addition to detection<br />
per<strong>for</strong>mance.<br />
From the results it is evident that the measured FM Broadcast signals due to the randomness in<br />
their modulation, exhibit noise like characteristics and behaviour. This is observed as the<br />
ambiguity function plots can be approximated by the ideal thumbtack ambiguity function,<br />
there<strong>for</strong>e providing the radar with unambiguous range and doppler in<strong>for</strong>mation. However, it is<br />
additionally clear that the ambiguity function depends largely on the modulating <strong>for</strong>mat.<br />
Thus there is a variation in per<strong>for</strong>mance between the pop music and dance music channels and<br />
the rock and classical music channels which exhibit a slightly degraded per<strong>for</strong>mance. Speech<br />
content shows the poorest per<strong>for</strong>mance though, as the per<strong>for</strong>mance does vary significantly with<br />
time, especially during pauses in words due to the low spectral content. This variation in<br />
per<strong>for</strong>mance does have a significant impact on the radar's potential detection ability and will be<br />
one that does vary with time, as seen from Equation 2.6 and Table A.6, this variation translates<br />
to a change in processing gain <strong>of</strong> 15dB.<br />
Table 5.2 summarises the ambiguity function per<strong>for</strong>mance <strong>of</strong> the measured signals.<br />
Signal<br />
Table 5.2: Summary <strong>of</strong> wave<strong>for</strong>m ambiguity function per<strong>for</strong>mances<br />
Range Resolution<br />
(km)<br />
Effective<br />
Bandwidth<br />
(kHz)<br />
Peak range<br />
Sidelobe level<br />
(dB)<br />
Peak doppler<br />
Sidelobe level<br />
(dB)<br />
RAD5 3.38 44.3 - 30 - 27<br />
<strong>Radio</strong> 2000 7.94 18.9 - 18 - 34<br />
Classic FM 7.98 18.8 - 30 - 33<br />
Umhlobe 3.33 45.0 - 31 - 25<br />
RGHP 2.36 63.6 - 30 - 37<br />
RSGR 9.32 16.1 - 18 - 36<br />
SAFM 7.54 19.9 - 20 - 40<br />
Average 5.98 32.37 -25.29 -33.14<br />
17
Chapter 6:<br />
A limiting factor in the per<strong>for</strong>mance <strong>of</strong> radar applications is the ability <strong>of</strong> the radar's frequency<br />
reference to maintain timing accuracy. Chapter 6 thus is a presentation <strong>of</strong> experimental<br />
measurements aimed at determining the stability <strong>of</strong> the local oscillators present in the TVRx<br />
frontend daughterboards <strong>of</strong> the USRP and hence the PCL receiver.<br />
Figure 1.4: USRP with two TVRx boards, reference channel and target channel<br />
The local oscillator in the TVRx daughterboard <strong>of</strong> the USRP is a 4MHz quartz crystal oscillator.<br />
It is <strong>for</strong> this reason that the properties <strong>of</strong> crystal oscillators and the associated per<strong>for</strong>mance<br />
parameters <strong>of</strong> interest were investigated. These parameters include accuracy, reproducibility and<br />
stability. It was found that crystal oscillators provide good per<strong>for</strong>mance at a reasonable price and<br />
dominate the field <strong>of</strong> frequency sources [19].<br />
Furthermore, the techniques associated with the specification <strong>of</strong> stability were presented,<br />
including measurement <strong>of</strong> the frequency deviation and the Allan variance. These tests concluded<br />
that the crystal ageing effect is reduced over the measurement period <strong>of</strong> thirty minutes, showing<br />
a variation from 2.3 to 0.8 parts per 10 million. In order to qualify the results <strong>of</strong> this chapter a<br />
5% error in the calculated doppler shift is set as the maximum acceptable limit <strong>of</strong> frequency<br />
shift. Thus a frequency drift <strong>of</strong> no greater than 5Hz is set as the requirement <strong>of</strong> the system.<br />
Chapter 7:<br />
This chapter presents the conclusions and recommendations <strong>of</strong> future work.<br />
In particular the ambiguity function is examined and the per<strong>for</strong>mance <strong>of</strong> commercial FM radio<br />
broadcasts as the radar wave<strong>for</strong>m is determined and summarised by Table 5.2, reproduced here.<br />
18
Signal<br />
Table 5.2: Summary <strong>of</strong> wave<strong>for</strong>m ambiguity function<br />
Range Resolution<br />
(km)<br />
Effective<br />
Bandwidth<br />
(kHz)<br />
Peak range<br />
Sidelobe level<br />
(dB)<br />
Peak doppler<br />
Sidelobe level<br />
(dB)<br />
RAD5 3.38 44.3 - 30 - 27<br />
<strong>Radio</strong> 2000 7.94 18.9 - 18 - 34<br />
Classic FM 7.98 18.8 - 30 - 33<br />
Umhlobe 3.33 45.0 - 31 - 25<br />
RGHP 2.36 63.6 - 30 - 37<br />
RSGR 9.32 16.1 - 18 - 36<br />
SAFM 7.54 19.9 - 20 - 40<br />
Average 5.98 32.37 -25.29 -33.14<br />
The analysis <strong>of</strong> the FM broadcast signal ambiguity plots reveal the attractive per<strong>for</strong>mance <strong>of</strong> the<br />
signals as they in general tend toward the idealised thumbtack. However, this is greatly<br />
dependant upon the instantaneous modulating content, revealing highly degraded per<strong>for</strong>mance in<br />
the case <strong>of</strong> a signal with low spectral content such as speech with many pauses and breaks.<br />
Furthermore, the SDR paradigm and technology is examined, with discussion around the design<br />
considerations. The USRP, the TVRx daughterboard and GNU<strong>Radio</strong> are examined further as a<br />
potential receiver and development environment, in this light.<br />
GNU<strong>Radio</strong>, is found to be a powerful toolkit that implements numerous complex signal<br />
processing elements and additionally interfaces with the USRP that provides flexible and<br />
powerful computing capabilities. Thus combining to provide a powerful and widely versatile<br />
toolset <strong>for</strong> the development <strong>of</strong> a variety <strong>of</strong> radio devices.<br />
The system meets the low cost ambitions costing just over US$500.00 <strong>for</strong> the USRP<br />
motherboard and a single daughterboard. The receiver's frontend provides a static bandwidth <strong>of</strong><br />
6MHz and a tunable range between 50MHz and 800MHz. The noise characterisation <strong>of</strong> the<br />
receiver reveals a NF <strong>of</strong> 10dB, a sensitivity <strong>of</strong> -105dB and a dynamic range <strong>of</strong> 62dB.<br />
Finally, the investigation into the stability <strong>of</strong> the daughterboard frontend oscillators due to ageing<br />
effects is shown to be steady through examination <strong>of</strong> the fractional frequency variation as well as<br />
the Allan variance, showing a fractional frequency variation from 2.3 to 0.8 parts per 10 million<br />
and a frequency drift no greater than 4Hz. This falls within the maximum allowance <strong>of</strong> 5Hz,<br />
calculated in section 6.1.<br />
Appendix A:<br />
Ambiguity Analysis Results.<br />
Appendix B:<br />
Relevant datasheets and Sentech transmission table.<br />
Bibliography<br />
19
Chapter 2<br />
Passive Coherent Location<br />
Conventional (dedicated) radar systems are comprised <strong>of</strong> a dedicated transmitter and receiver.<br />
In a passive radar system, however, there is no dedicated transmitter. Instead the receiver uses<br />
third-party non cooperative transmitters in the environment. PCL systems are bistatic in nature<br />
and thus measure the difference in time <strong>of</strong> arrival between the signal directly from the<br />
transmitter and the signal reflected <strong>of</strong>f the target. This allows the bistatic range <strong>of</strong> the object to<br />
be determined. In addition to the bistatic range, in order <strong>for</strong> the the location, direction and<br />
velocity <strong>of</strong> the target to be calculated the bistatic doppler shift <strong>of</strong> the reflected signal and its<br />
direction <strong>of</strong> arrival is determined.<br />
Clearly, in a dedicated radar system, the the transmitter and its location are directly under the<br />
control <strong>of</strong> the radar engineer, hence timing in<strong>for</strong>mation related to the transmission <strong>of</strong> the pulse<br />
and the <strong>for</strong>m <strong>of</strong> the transmitted wave<strong>for</strong>m are exactly known. This allows the target range to be<br />
easily calculated and <strong>for</strong> a matched filter to be used to achieve an optimal signal to noise ratio<br />
(SNR) in the receiver.<br />
A passive radar however, does not have control over the transmitter, its location, the nature <strong>of</strong><br />
the transmitted wave<strong>for</strong>m or timing in<strong>for</strong>mation related to the transmission and there<strong>for</strong>e must<br />
use a dedicated receiver “reference channel” to monitor each transmitter being exploited.<br />
2.1 <strong>System</strong> level description <strong>of</strong> PCL<br />
The purpose <strong>of</strong> this section is to provide the reader with insight into the workings and<br />
implementation <strong>of</strong> the overall system, without providing an overwhelming detail <strong>of</strong> theory. This<br />
will enable the reader to understand the context within which the later sections fit.<br />
20
A passive radar would typically consist <strong>of</strong> the following processing steps:[33]. This is illustrated<br />
in Figure 2.1<br />
●<br />
●<br />
●<br />
●<br />
●<br />
●<br />
●<br />
●<br />
Reception <strong>of</strong> the direct signal from the transmitter(s) and from the surveillance region on<br />
dedicated low-noise, linear, digital receivers.<br />
Digital beam <strong>for</strong>ming to determine the DOA <strong>of</strong> signals and spatial rejection <strong>of</strong> strong inband<br />
interference.<br />
Adaptive filtering to cancel any unwanted direct signal returns in the surveillance<br />
channel(s).<br />
Transmitter-specific signal conditioning.<br />
Cross-correlation <strong>of</strong> Doppler-shifted copies <strong>of</strong> the reference channel with the surveillance<br />
channel(s) to determine target bistatic range and Doppler.<br />
Detection using a constant false alarm rate (CFAR) scheme.<br />
Association and tracking <strong>of</strong> object returns in range/doppler space, known as “trajectory<br />
tracking” typically employing a Kalman filter.<br />
Association and fusion <strong>of</strong> line tracks from each transmitter to <strong>for</strong>m the final estimate <strong>of</strong><br />
an object’s location, heading and speed.<br />
21
Figure 2.1: PCL system diagram [33]<br />
2.1.1 Receiver system<br />
It is essential that the PCL receiver should have a low noise figure, high dynamic range and high<br />
linearity. This is as a result <strong>of</strong> the nature <strong>of</strong> PCL systems using illuminators <strong>of</strong> opportunity with<br />
continuous wave signals, wherein they must detect very small target returns in the presence <strong>of</strong><br />
continuous and significant interference.<br />
In order to achieve the processing gain necessary to detect these weak target returns in a<br />
background <strong>of</strong> noise and interference it is necessary to achieve an equivalent to the optimal<br />
matched filter processing used in conventional radar systems.<br />
22
Passing target echoes through a matched filter is equivalent to the correlation <strong>of</strong> the radar echo<br />
with a delayed replica <strong>of</strong> the transmitted signal which is obtained via the reference channel<br />
mentioned earlier.<br />
The greatest source <strong>of</strong> interference and hence limitation on the system per<strong>for</strong>mance is the Direct<br />
Path Interference (DPI) received from the transmitter being used to detect aircraft. This<br />
unwanted DPI signal correlates perfectly with the reference signal and produces range and<br />
doppler sidelobes that are several orders <strong>of</strong> magnitude greater than the echoes that are sought<br />
[18].<br />
To detect anything but the closest <strong>of</strong> targets, it is necessary to remove this signal,by both angular<br />
nulling with the antenna and adaptive echo cancellation in the receiver. These strategies are<br />
further elucidated in the sections to follow.<br />
However, eventually the dynamic range <strong>of</strong> the receiver limits the cancellation and so the<br />
principal limitation on system per<strong>for</strong>mance lies with the analogue to digital converter (ADC)<br />
technology [16].<br />
Further interference is due co-channel interference coming from other transmissions operating at<br />
the same frequency within a single frequency network. This has a similar effect to the DPI.<br />
In addition, the signal is always received against a background <strong>of</strong> reflections from the surface,<br />
i.e. clutter, and could be buried under these interferences. Their power density at the radar<br />
antenna, in addition to correlation between the desired signal and the interferences, further<br />
influences the system’s per<strong>for</strong>mance.<br />
These effects highlight the importance <strong>of</strong> conducting a characterisation <strong>of</strong> the USRP receiver.<br />
A more thorough treatment <strong>of</strong> these interferences and the methods used to mitigate them are<br />
discussed in [18] [12] [13].<br />
2.1.2 Digital beam<strong>for</strong>ming<br />
Passive radar systems use antenna arrays with several antenna elements and element-level<br />
digitisation. This allows the direction <strong>of</strong> arrival <strong>of</strong> echoes to be calculated using standard radar<br />
beam<strong>for</strong>ming techniques, such as amplitude monopulse using a series <strong>of</strong> fixed, overlapping<br />
beams or more sophisticated adaptive beam<strong>for</strong>ming. Alternatively, some research systems have<br />
used only a pair <strong>of</strong> antenna elements and the phase-difference <strong>of</strong> arrival to calculate the direction<br />
<strong>of</strong> arrival <strong>of</strong> the echoes (known as phase interferometry) [15] [40].<br />
2.1.3 Signal conditioning<br />
Adaptive nulling is spatial filtering through an adaptive beam<strong>for</strong>mer (ADBF), to null the<br />
interference. To null the interference, the output power <strong>of</strong> the beam<strong>for</strong>mer along the direction <strong>of</strong><br />
the interference must be minimized, whereas the output power along the desired signal’s<br />
direction must be maximized. This may include high quality analogue bandpass filtering <strong>of</strong> the<br />
signal, channel equalization to improve the quality <strong>of</strong> the reference signal, removal <strong>of</strong> unwanted<br />
structures in digital signals to improve the radar ambiguity function or even complete<br />
reconstruction <strong>of</strong> the reference signal from the received digital signal [18] [33].<br />
23
2.1.4 Adaptive filtering<br />
The principal limitation in detection range <strong>for</strong> most passive radar systems is the signal to<br />
interference ratio (SIR), due to the large and constant direct signal received from the transmitter.<br />
To remove this, an adaptive filter can be used to remove the direct signal in a process similar to<br />
active noise control. This step is essential to ensure that the range/doppler sidelobes <strong>of</strong> the direct<br />
signal do not mask the smaller echoes in the subsequent cross-correlation stage.<br />
In a few specific cases, the direct interference is not a limiting factor, due to the transmitter being<br />
beyond the horizon or obscured by terrain (such as with the Manastash Ridge Radar), but this is<br />
the exception rather than the rule, as the transmitter must normally be within line-<strong>of</strong>-sight <strong>of</strong> the<br />
receiver to ensure good low-level coverage [33].<br />
2.1.5 Cross-correlation processing<br />
The key processing step in a passive radar is cross-correlation. This step acts as the matched<br />
filter to provide the necessary processing gain thereby maximising the SNR. The response to this<br />
matched filter with respect to time and doppler shift <strong>of</strong> the carrier frequency from a normalized<br />
version <strong>of</strong> the transmitted wave<strong>for</strong>m is known as the ambiguity function.<br />
This topic is treated with greater detail in section 2.6 <strong>of</strong> this chapter.<br />
Most analogue and digital broadcast signals are noise-like in nature, and as a consequence they<br />
tend to only correlate with themselves. This presents a problem with moving targets, as the<br />
doppler shift imposed on the echo means that it will not correlate with the direct signal from the<br />
transmitter. As a result, the cross-correlation processing must implement a bank <strong>of</strong> matched<br />
filters, each matched to a different target doppler shift.<br />
2.1.6 Target detection<br />
Targets are detected on the cross-correlation surface by applying an adaptive threshold, and<br />
declaring all returns above this surface to be targets. A standard cell-averaging CFAR algorithm<br />
is typically used and determines the range and doppler <strong>of</strong> each target.<br />
2.1.7 Trajectory tracking<br />
At this stage in the processing the system has determined the range and doppler <strong>of</strong> a number <strong>of</strong><br />
targets. In order to further process the data it is necessary to associate this plot data with<br />
individual targets and this is per<strong>for</strong>med using a conventional Kalman filter. Most false alarms are<br />
rejected during this stage <strong>of</strong> the processing.<br />
2.1.8 Track association and state estimation<br />
After having associated plots-to-targets, the range and doppler data <strong>for</strong> each target are processed<br />
by a non-linear estimator to determine the target’s location, speed and heading. Use <strong>of</strong> a nonlinear<br />
estimator allows optimum use <strong>of</strong> the doppler in<strong>for</strong>mation in this tracking process.<br />
24
2.1.9 Narrow band and CW illumination sources<br />
The above description assumes that the wave<strong>for</strong>m <strong>of</strong> the transmitter being exploited possesses a<br />
usable radar ambiguity function and hence cross-correlation yields a useful result. Some<br />
broadcast signals, such as analogue television, contain a structure in the time domain that yields<br />
a highly ambiguous or inaccurate result when cross-correlated. In this case, the processing<br />
described above is ineffective.<br />
If the signal contains a continuous wave (CW) component, however, such as a strong carrier<br />
tone, then it is possible to detect and track targets in an alternative way. Over time, moving<br />
targets will impose a changing doppler shift and direction <strong>of</strong> arrival on the CW tone that is<br />
characteristic <strong>of</strong> the location, speed and heading <strong>of</strong> the target. It is there<strong>for</strong>e possible to use a<br />
non-linear estimator to estimate the state the <strong>of</strong> the target from the time history <strong>of</strong> the doppler<br />
and bearing measurements.<br />
The vision carrier <strong>of</strong> analogue TV signals has been shown to have successfully achieved this<br />
[15]. However, target track initiation is slow and difficult, and so the use <strong>of</strong> narrow band signals<br />
i.e. FM radio broadcasts is probably best considered as an ancillary to the use <strong>of</strong> illuminators<br />
with better ambiguity surfaces. Later sections examine the merits <strong>of</strong> such a choice in greater<br />
detail.<br />
2.2 Advantages and Disadvantages <strong>of</strong> PCL <strong>System</strong>s<br />
PCL systems operate in bistatic and multistatic configurations, thus inherit all the associated<br />
advantages and disadvantages.<br />
Further additional advantages include:<br />
●<br />
●<br />
●<br />
●<br />
Low procurement costs as they do not have any transmitter hardware.<br />
Lower costs <strong>of</strong> operation and maintenance due to lack <strong>of</strong> transmitter and moving parts.<br />
Covert operation, including no need <strong>for</strong> frequency allocations.<br />
Physically small and hence easily deployed in places where conventional radars cannot<br />
be.<br />
● Rapid updates, typically once a second [33].<br />
Disadvantages:<br />
●<br />
●<br />
Lack <strong>of</strong> control over the <strong>for</strong>m, nature and origin <strong>of</strong> the transmitters and signal wave<strong>for</strong>m.<br />
Per<strong>for</strong>mance limitations and complexity <strong>of</strong> deployment due to direct path interference,<br />
co-channel interference, etc.<br />
25
2.3 History and Examples <strong>of</strong> PCL <strong>System</strong>s<br />
Much <strong>of</strong> the early work published on PCL systems was conducted at University College London<br />
(UCL) and was undertaken in the late 1970s.<br />
A number <strong>of</strong> experimental bistatic systems have been built and evaluated by Schoenenberger<br />
who designed and built a system using a UHF Air Traffic Control radar at Heathrow airport as<br />
an illuminator, and investigated particularly the problems <strong>of</strong> synchronization between receiver<br />
and transmitter [28]. A real-time co-ordinate correction scheme was also developed <strong>for</strong> this<br />
system. Further developments included a digital beam<strong>for</strong>ming array <strong>for</strong> pulse chasing<br />
experiments and a coherent MTI system using clutter from stable local echoes as a phase<br />
reference [4].<br />
Subsequent work at UCL attempted to use UHF television transmissions as illuminators <strong>of</strong><br />
opportunity, to detect aircraft targets landing and taking <strong>of</strong>f from Heathrow airport.<br />
This work made use <strong>of</strong> the pulsed-like nature <strong>of</strong> parts <strong>of</strong> the television wave<strong>for</strong>m <strong>for</strong> bistatic use<br />
and showed the ability to receive clutter from the surrounding buildings. By implementing a two<br />
pulse MTI canceller, <strong>of</strong>f line, they were able to “track” moving targets. This system was<br />
impractical in the sense that they required a special transmission wave<strong>for</strong>m and only has a range<br />
resolution <strong>of</strong> 1800m and range ambiguity <strong>of</strong> 9600m.<br />
In attempt to improve the system per<strong>for</strong>mance, the use <strong>of</strong> a multi burst test pattern was<br />
investigated. This once again yielded positive results <strong>for</strong> static objects, yet failed to resolve<br />
moving targets.<br />
This work illustrated that terrestrial television, in the time domain, was not suitable <strong>for</strong> use in a<br />
bistatic configuration and that the use <strong>of</strong> a pulse radar transmitter is the best case, however when<br />
the transmitter is not radar like, then the autocorrelation function <strong>of</strong> the transmitted wave<strong>for</strong>m is<br />
<strong>of</strong> paramount importance [15].<br />
In addition to this, the transmit power should be proportionate to the coverage required. In the<br />
case <strong>of</strong> complex modulation functions such as television, the calculation is made on the basis <strong>of</strong><br />
the power <strong>of</strong> that part <strong>of</strong> the spectrum used by the radar. Moreover, the radiation pattern <strong>of</strong> the<br />
transmitter should be either omnidirectional or pencil-beam. Finally, the modulation bandwidth<br />
must be commensurable with the required range resolution [12] .<br />
Further work on television in the time domain using sophisticated correlation techniques which<br />
were applied to Direct Broadcast by Satellite (DBS) TV signal was still not able to detect<br />
moving targets at useful range. Thus suggesting that the problems associated with useful target<br />
detection using DBS TV prevent it from being a practical approach to detecting airborne<br />
targets[8].<br />
Howland [15] developed a UHF <strong>for</strong>ward scatter system based on television transmissions. A<br />
<strong>for</strong>ward scatter system is not able to provide range in<strong>for</strong>mation, due to the nature <strong>of</strong> the bistatic<br />
geometry, Thus a different approach was adopted, measuring direction <strong>of</strong> arrival (DOA) by<br />
phase interferometry and doppler shift <strong>of</strong> the vision carrier <strong>of</strong> the television signal. The angle <strong>of</strong><br />
arrival <strong>of</strong> a target echo is related to the phase difference <strong>of</strong> the received signal at the surveillance<br />
antennas by:<br />
= 2 d<br />
(2.1)<br />
<br />
sin<br />
26
where:<br />
<br />
<br />
<br />
d<br />
is the angle <strong>of</strong> arrival <strong>of</strong> a target echo<br />
is the phase difference <strong>of</strong> the received signal<br />
is the wavelength <strong>of</strong> the signal<br />
is the distance between the dipoles<br />
This work focused on developing signal processing techniques used <strong>for</strong> target tracking by means<br />
<strong>of</strong> an extended Kalman filter algorithm and assumes that there is only one receiver system, and<br />
one remotely located television transmitter. On its own the in<strong>for</strong>mation extracted from a single<br />
measurement <strong>of</strong> doppler shift and DOA <strong>of</strong> the target echo, is not useful. There<strong>for</strong>e, a series <strong>of</strong><br />
measurements <strong>for</strong> doppler and DOA with respect to time were taken.<br />
This choice was due to the unique association <strong>of</strong> the target echo's change in doppler shift and<br />
DOA with its course and velocity. Howland's research bore similarity to the work done at UCL,<br />
with the exception that his work was done in the frequency domain and exploited DOA and was<br />
able to demonstrate tracking <strong>of</strong> aircraft targets at ranges in excess <strong>of</strong> 100km from the receiver.<br />
Further work by Howland [16] investigated the use <strong>of</strong> FM radio broadcast transmissions, which<br />
closely resembles his previous work based on television transmissions, employing doppler, range<br />
and bearing data. Adaptive filtering was applied to two surveillance channels to reject the direct<br />
transmitter signal interference. A conventional CFAR detection scheme was applied to range and<br />
doppler in<strong>for</strong>mation resulting from the matched filtering to determine the range and doppler <strong>of</strong><br />
each target. With only two surveillance channels the direction finding system again uses phase<br />
interferometry to estimate the target bearing.<br />
Target tracking was employed by implementation <strong>of</strong> a Kalman filter and nonlinear estimation to<br />
determine the target’s location, speed and heading. Use <strong>of</strong> a nonlinear estimator allowed<br />
optimum use <strong>of</strong> the doppler in<strong>for</strong>mation in this tracking process. Via this approach Howland was<br />
able to demonstrate tracking <strong>of</strong> aircraft at ranges in excess <strong>of</strong> 150km from the receiver.<br />
The Manastash Ridge Radar is a system conceived and built by John Sahr <strong>of</strong> the University <strong>of</strong><br />
Washington, Seattle, <strong>for</strong> studies <strong>of</strong> the ionosphere [26]. It uses a single VHF radio transmitter as<br />
illuminator, and a receiver separated from the transmitter by a large mountain range<br />
(Mt.Rainier). The receiving system is based on a standard digitizer card and PC, and is<br />
extremely simple and cheap, approximately US$15 000. Synchronization is achieved by GPS,<br />
giving uncertainties <strong>of</strong> 100ns in timing which equates to 15 m in range and 0.01 Hz in doppler.<br />
Although the purpose <strong>of</strong> the system is <strong>for</strong> ionospheric studies, it does detect aircraft targets at<br />
ranges up to about 100km. Their system vividly demonstrates that high per<strong>for</strong>mance can be<br />
achieved from simple and inexpensive PCL systems.<br />
Silent Sentry developed by the Lockheed Martin company, based on multiple VHF FM radio and<br />
television transmissions. It has demonstrated tracking <strong>of</strong> aircraft and space targets renewing<br />
tracking targets can be completed eight times in a second. The database <strong>of</strong> the system has stored<br />
in it, about 55,000 correlative FM broadcast and digital audio broadcast parameters and is<br />
advertised as being applicable to: air surveillance and tracking in areas <strong>of</strong> limited coverage;<br />
capable <strong>of</strong> tracking low flying, non-cooperative, slow moving targets; continuous total volume<br />
surveillance <strong>of</strong> air breathing and ballistic objects; low acquisition and operations cost,<br />
unattended remotely managed.<br />
27
2.4 Applications <strong>of</strong> PCL<br />
Reported applications <strong>of</strong> PCL include:[10]<br />
●<br />
air-space surveillance<br />
● maritime surveillance<br />
●<br />
●<br />
●<br />
●<br />
atmospheric studies and ionospheric studies<br />
oceanography<br />
mapping lightning channels in thunderstorms<br />
monitoring radioactive pollution<br />
There are also reports <strong>of</strong> algorithm development <strong>for</strong> interferometry, target tracking and target<br />
classification [10]. This panoramic diversity <strong>of</strong> systems and applications is indicative <strong>of</strong> the<br />
increasing importance <strong>of</strong> PCL systems.<br />
2.5 Typical Illuminators<br />
A wide variety <strong>of</strong> RF emissions in the <strong>for</strong>m <strong>of</strong> TV and radio broadcasts in addition to terrestrial<br />
and space based communications has resulted in a wide range <strong>of</strong> signal types available <strong>for</strong><br />
exploitation by passive radar. Furthermore, many such transmissions are at VHF and UHF<br />
frequencies, which allows these parts <strong>of</strong> the spectrum not normally available <strong>for</strong> radar use, and at<br />
which stealth treatment <strong>of</strong> targets may be less effective, to be used.<br />
Typical broadcast service sources include:<br />
●<br />
●<br />
●<br />
●<br />
●<br />
●<br />
Analogue TV signals<br />
FM broadcast radio signals<br />
Global <strong>System</strong> <strong>for</strong> Mobile (GSM) base stations<br />
Digital audio broadcasts<br />
Digital video broadcasts<br />
Terrestrial HDTV<br />
So Why FM radio<br />
Of all the transmitters <strong>of</strong> opportunity available in the environment, the typical broadcast service<br />
sources represent some <strong>of</strong> the most attractive <strong>for</strong> surveillance purposes, owing to their inherent<br />
attractive properties <strong>of</strong> very broad coverage and relatively high transmitter powers.<br />
28
Satellite signals have generally been found to be inadequate <strong>for</strong> passive radar use [33]. This is as<br />
a result <strong>of</strong> their transmission powers being too low, or because the orbits <strong>of</strong> the satellites are such<br />
that illumination is too infrequent. However, exceptions to this include the exploitation <strong>of</strong><br />
satellite based radar and satellite radio systems.<br />
Digital audio transmitters emit signals at a much higher frequency than FM broadcast<br />
transmissions, however FM broadcast transmissions are comparatively higher in their power.<br />
Thus, the maximum detection ranges <strong>of</strong>fered by FM broadcast transmissions are greater than that<br />
<strong>of</strong> Digital audio [10]. Furthermore, the coverage <strong>of</strong> Digital audio is generally poor, although the<br />
transmitter networks are expanding elsewhere, they are currently non-existent in South Africa.<br />
Cellular phone base stations transmit at a rather low power and would seem there<strong>for</strong>e to be<br />
limited in their per<strong>for</strong>mance and application. However, there is an extensive network and targets<br />
could be tracked through such a network and hence the coverage may be extended greatly. This<br />
however comes at the cost <strong>of</strong> increased system complexity. Furthermore, base stations<br />
deliberately concentrate their emissions towards the ground and may not necessarily have good<br />
coverage <strong>of</strong> higher altitude aircraft.<br />
Analogue television transmitters, with high radiated powers and a pulse like wave<strong>for</strong>m structure<br />
present themselves as an obvious choice <strong>of</strong> illuminator. However, in spite <strong>of</strong> these instant points<br />
<strong>of</strong> appeal, it has been shown that the wave<strong>for</strong>m is far from suited <strong>for</strong> radar usage when used in a<br />
conventional radar matched filtering approach [12]. In an alternative approach however, making<br />
use <strong>of</strong> doppler and bearing in<strong>for</strong>mation in echoes <strong>of</strong> the television video carrier it is possible to<br />
track aircraft at ranges <strong>of</strong> up to 260km from the receiver and 150km from the transmitter [15].<br />
This approach is referred to as narrow band processing as only a few kilohertz and thus small<br />
percentage <strong>of</strong> the 5.5MHz television wave<strong>for</strong>m bandwidth is processed to determine the target<br />
doppler shifts. The disadvantages <strong>of</strong> this stem from the relatively low in<strong>for</strong>mation content in the<br />
doppler measurements and the system must observe the target’s doppler history <strong>for</strong> an extended<br />
time be<strong>for</strong>e there is sufficient in<strong>for</strong>mation to locate the target.<br />
Moreover, the use <strong>of</strong> non-linear estimation techniques to calculate the target’s trajectory mean<br />
that a good estimate <strong>of</strong> the target’s location must be initially available. In the case <strong>of</strong> a <strong>for</strong>ward<br />
scatter radar the unique geometry can be manipulated in order to determine an expression <strong>of</strong> the<br />
target’s location [2].<br />
In the general bistatic problem it is necessary employ global optimisation schemes [15]. The<br />
<strong>for</strong>mer approach is limited in its operational applications and the latter is neither robust nor<br />
computationally efficient.<br />
FM broadcast transmissions have been used to probe the ionosphere [26] and there<strong>for</strong>e might be<br />
expected to have useful height surveillance. Their high powers and good coverage make FM<br />
broadcast radio transmissions particularly well suited to air target detection <strong>for</strong> both civil and<br />
military applications. Moreover, they could be used <strong>for</strong> marine navigation in coastal waters<br />
although clutter will be a more significant problem.<br />
In PCL systems wideband processing is defined as the use <strong>of</strong> a receiver that has a bandwidth that<br />
is comparable to that <strong>of</strong> the wave<strong>for</strong>m being processed. A typical FM radio broadcast occupies a<br />
bandwidth <strong>of</strong> approximately 100kHz, thus <strong>of</strong>fering a potential range resolution <strong>of</strong> up to 1500m.<br />
It is clear that the signal <strong>of</strong>fers useful target ranging in<strong>for</strong>mation. However, as a result <strong>of</strong> this<br />
lower bandwidth available, range and bearing are a factor <strong>of</strong> ten or so worse than a conventional<br />
microwave radar.<br />
29
Doppler, on the other hand, is two or three orders <strong>of</strong> magnitude more accurate. This is due to the<br />
extended integration times that are potentially achieved by passive radar. The doppler<br />
in<strong>for</strong>mation can be used to provide a resolution comparable to conventional radars and by<br />
simultaneously using multiple transmitters the system can achieve target location accuracies that<br />
may be even better.<br />
However, the suitability <strong>of</strong> a signal <strong>for</strong> target location is governed by more than its bandwidth.<br />
Of greater value and importance is the ability <strong>of</strong> the radar receiver to locate the target<br />
unambiguously. For this we must compute its ambiguity function.<br />
2.6 Relevant Equations<br />
2.6.1 The Bistatic Radar equation<br />
The starting point <strong>of</strong> an analysis <strong>of</strong> the per<strong>for</strong>mance <strong>of</strong> a system is the radar equation. The radar<br />
equation <strong>for</strong> a bistatic system is derived in the same way as <strong>for</strong> monostatic systems.<br />
P r<br />
P n<br />
= P t G t<br />
4 r 1<br />
2 ⋅ b ⋅ 1<br />
4 r 2<br />
2 ⋅G r 2<br />
4 ⋅ 1<br />
k T 0<br />
B F ⋅L<br />
(2.2)<br />
where: [10]<br />
P r<br />
P n<br />
P t<br />
G t<br />
G r<br />
r 1<br />
r 2<br />
b<br />
<br />
k<br />
T 0<br />
B<br />
F<br />
L ≤1<br />
is the received signal power<br />
is the receiver noise power<br />
is the transmit power<br />
is the transmit antenna gain<br />
is the receive antenna gain<br />
is the transmitter to target range<br />
is the target to receiver range<br />
is the bistatic radar cross section<br />
is the signal wavelength<br />
is Boltzmann's constant<br />
is the noise reference temperature, 290 K (standard room temperature)<br />
is the receiver effective bandwidth<br />
is the receiver effective noise figure<br />
are system losses<br />
30
setting the values:<br />
●<br />
●<br />
b<br />
= m<br />
r 1<br />
= r 2<br />
= r m<br />
where:<br />
m<br />
is the monostatic radar cross section<br />
r m<br />
is the radar to target range<br />
allows <strong>for</strong> the reduction <strong>of</strong> the bistatic range equation to the monostatic case.<br />
In order to use the radar range equation appropriately and correctly to predict per<strong>for</strong>mance <strong>of</strong> a<br />
radar system it is necessary to understand the correct value <strong>of</strong> each parameter to be used.<br />
The transmit power P t <strong>of</strong> the various illuminators <strong>of</strong> opportunity exploited in PCL is high.<br />
This is due to the fact that the intended receivers <strong>of</strong>ten have inefficient antennas and poor noise<br />
figures, with the transmission paths far from line <strong>of</strong> sight. Thus the substantial transmit power is<br />
necessary to overcome the inefficiencies. However, <strong>for</strong> example, the analogue TV broadcast has<br />
pronounced ambiguities every 64 s and thus does not exhibit favourable ambiguity<br />
properties and a more attractive ambiguity may be realised by using a portion <strong>of</strong> the signal<br />
spectrum, at the expense <strong>of</strong> reduced power. There<strong>for</strong>e, when employing the radar equation it is<br />
necessary to consider only the portion <strong>of</strong> the signal spectrum that is used <strong>for</strong> the radar purposes.<br />
This may not necessarily be the same as the power <strong>of</strong> the entire signal spectrum.<br />
2.6.2 Bistatic Radar Cross Section<br />
The likelihood <strong>of</strong> target detection and location is dependant on the spatially influenced bistatic<br />
RCS and target dynamics, in addition to the radar design parameters. The use <strong>of</strong> conventional<br />
processing techniques will result in the detection <strong>of</strong> targets in range, doppler and angle.<br />
Relatively little has been published regarding the bistatic RCS <strong>of</strong> targets and this remains an area<br />
<strong>for</strong> future research. However, early work, that resulted in the <strong>for</strong>mulation <strong>of</strong> the bistatic<br />
equivalence theorem, states that that the bistatic RCS is equal to the monostatic RCS at the<br />
bisector <strong>of</strong> the bistatic angle β, reduced in frequency by the factor cos (β/2)[9].<br />
This theorem however, is valid typically, <strong>for</strong> small angles, when the bistatic angle, β, is 5 O or<br />
less. There<strong>for</strong>e, above 5 O the target bistatic RCS will not in general be the same as the<br />
monostatic RCS, although will be comparable <strong>for</strong> targets that have a low monostatic RCS and<br />
due to shadowing that occurs in the monostatic geometry but not the bistatic geometry. In a<br />
monostatic radar, phase interference from two or more targets causes a distortion <strong>of</strong> the echo<br />
signal. This in turn results in the apparent phase centre <strong>of</strong> the radar reflection swinging between<br />
the targets. This random movement <strong>of</strong> radar reflecting centres, leads to jittered angle tracking,<br />
known as target glint [14]. The effect <strong>of</strong> target glint is reduced in the bistatic RCS region, thus<br />
the bistatic RCS has the advantage <strong>of</strong> glint reduction.<br />
As the bistatic angle in increased to 180 O the <strong>for</strong>ward scatter region is encountered where the<br />
target lies on the transmitter-receiver baseline. Within this region, the bistatic RCS can be many<br />
times greater than that <strong>of</strong> the monostatic RCS. The magnitude <strong>of</strong> the target’s <strong>for</strong>ward scatter<br />
return does not depend on the material composition, there<strong>for</strong>e irregular shaped targets, and radar<br />
absorbent materials used on targets are still able to be detected.<br />
31
Any target illuminated by the transmitter on the baseline, when the targets dimensions are larger<br />
than the transmitted wavelength, produces a shadow. This shadow region occurs on the opposite<br />
side <strong>of</strong> the target from the transmitter. This maybe understood with reference to Babinet’s<br />
Principle. Babinet’s principle is an approximation according to which the amplitude <strong>of</strong> near<br />
<strong>for</strong>ward scattering by an opaque, planar object is the same as that <strong>of</strong> an aperture <strong>of</strong> the same<br />
shape and size in a perfectly conducting screen [10].<br />
The <strong>for</strong>ward scatter cross-section <strong>of</strong> a target with cross sectional area A is given by:<br />
b<br />
= 4 A2<br />
2<br />
(2.3)<br />
Where the radiation wavelength, is assumed small compared to the target dimension.<br />
The angular width <strong>of</strong> the scattered signal in the horizontal or vertical plane is given by:<br />
b<br />
= d<br />
(2.4)<br />
where:<br />
d is the target linear dimension.<br />
The dependence <strong>of</strong> b and b <strong>for</strong> a target <strong>of</strong> the size <strong>of</strong> a typical aircraft, is shown below.<br />
Where A = 10m 2 and d = 20m.<br />
Figure 2.2: Variation <strong>of</strong> RCS and angular width <strong>of</strong> a <strong>for</strong>ward scatter target against frequency [10]<br />
32
Angular width decreases with an increase in frequency. Thus the <strong>for</strong>ward scatter illumination is<br />
focused into an increasingly narrower beam. This implies that frequencies around VHF and UHF<br />
are more likely to be optimum <strong>for</strong> exploiting <strong>for</strong>ward scatter as target detection may be achieved<br />
over an adequately wide angular range. Taking advantage <strong>of</strong> the <strong>for</strong>ward scatter configuration,<br />
however, comes at the expense <strong>of</strong> target doppler and position in<strong>for</strong>mation. However, target<br />
location can be estimated using a combination <strong>of</strong> doppler and bearing as explained in [15].<br />
2.6.3 Bistatic Clutter<br />
The received signal at any location is the vector RF sum <strong>of</strong> the direct signal and <strong>of</strong> the multipath<br />
propagation [12]. Most <strong>of</strong> these multipath signal energies would result from reflection <strong>of</strong>f fixed<br />
objects, such as, buildings, other large objects, other aircraft, etc. This multipath propagation is<br />
defined as clutter. Clutter, in general, is a cause <strong>of</strong> degradation in radar system per<strong>for</strong>mance, as<br />
the target returns must compete with the clutter returns at the receiver. More <strong>of</strong>ten than not<br />
clutter RCS are larger than target RCS [4].<br />
Bistatic clutter is subject to greater variability than the monostatic case, because there are more<br />
variables associated with the geometry [34], in addition to frequency and surface composition.<br />
Relatively little work has been done in developing models <strong>for</strong> bistatic clutter and available<br />
experimental data from the intended receiver location at UCT is minimal,and thus remains an<br />
area worthy <strong>of</strong> future research.<br />
It is important to know the clutter power returns at the receiver, in order to prevent the receiver<br />
from saturating. Thus the work to be presented in chapter 4, characterising the receiver, will<br />
provide the reader with an indication <strong>of</strong> the expected per<strong>for</strong>mance in this regard.<br />
2.6.4 Integration Gain<br />
Radar signals are considered to be coherent when the relative phases <strong>of</strong> the signals have a known<br />
relationship, even though they may be separated in time. To improve the SNR a combination <strong>of</strong><br />
coherent and incoherent integration maybe implemented. Coherent integration is per<strong>for</strong>med by<br />
correlating the target echo with the reference signal and in essence, is the addition <strong>of</strong> a sequence<br />
<strong>of</strong> M coherent radar signals where the signals are summed vectorially and is regarded as a single<br />
wave<strong>for</strong>m. Through this process the SNR will increase by a factor M, as given by Equation 2.5, a<br />
derivation <strong>of</strong> this can be found in [24].<br />
SNR m<br />
= M × SNR 1<br />
(2.5)<br />
Where:<br />
M<br />
SNR m<br />
is the number <strong>of</strong> coherent radar signals.<br />
Is the the SNR <strong>of</strong> the average <strong>of</strong> M signals.<br />
In the case <strong>of</strong> incoherent integration, these M pulses will not add in phase and will not improve<br />
the SNR. Instead the signals are processed separately and the separate observations are then<br />
combined, usually by means <strong>of</strong> averaging.<br />
33
The relative roles <strong>of</strong> coherent and incoherent integration will be a function <strong>of</strong> the coherency <strong>of</strong><br />
the target echo and <strong>of</strong> competing noise and clutter signals.The maximum signal processing gain<br />
achievable is given by:<br />
G p<br />
= B T max<br />
(2.6)<br />
Where:<br />
B<br />
T max<br />
is the wave<strong>for</strong>m bandwidth.<br />
Is the maximum integration dwell time.<br />
The effective wave<strong>for</strong>m bandwidth <strong>of</strong> an FM radio transmission is known to vary with time,<br />
clearly this will translate to a variation in signal processing gain.<br />
A rule <strong>of</strong> thumb determining the maximum value <strong>for</strong> the integration dwell time or maximum<br />
coherent processing time is given by:<br />
T max =<br />
{ A R<br />
} 1/2<br />
(2.7)<br />
where:<br />
A R<br />
is the radial component <strong>of</strong> the target acceleration.<br />
The integration dwell time is the length <strong>of</strong> time taken to make an observation with a radar set,<br />
the integration dwell time is dependant on the wave<strong>for</strong>m coherence and target dynamics, namely<br />
target velocity and manoeuvrability.<br />
T max<br />
consequently, is additionally a time varying entity and care will need to be observed in<br />
its choice.<br />
2.6.5 Range Resolution and Maximum Unambiguous Range<br />
The minimum distance necessary to separate two targets in order to distinguish between them<br />
defines the range resolution.<br />
For both monostatic and bistatic cases, adequate separation between the two target echoes at the<br />
receiver is taken to be:<br />
c⋅<br />
(2.8)<br />
2<br />
where:<br />
<br />
is the compressed radar pulse width.<br />
34
Furthermore, (2.9)<br />
= 1 B<br />
and thus, Equation 2.8 can be restated as: c<br />
(2.10)<br />
2 B<br />
If the time delay between the echoes from two targets is > the pulse width then the two<br />
echoes are resolvable. However, if the targets are closer than their echoes merge and are<br />
unresolvable.<br />
In order to generate this separation at a bistatic receiver, two-point targets must lie on a bistatic<br />
isorange contour with a separation R B . In the monostatic case, the distance between the two<br />
circles are constant, unlike the bistatic case.<br />
The equation <strong>for</strong> a monostatic system resolution is given by the equation:<br />
R M<br />
= c⋅<br />
2<br />
(2.11)<br />
For a bistatic radar, separation between the ellipses depend on the bistatic angles. As the bistatic<br />
angle increases, so does the distance between the isorange contours, and vice versa. Eventually<br />
on an extended baseline, or long ranges, they become equispaced, and start to represent a<br />
monostatic case. An approximation <strong>for</strong> R B is:[40]<br />
R B<br />
= R M<br />
cos 2 <br />
(2.12)<br />
= c⋅<br />
2cos 2 <br />
The above equation is based on the monostatic case, reduced by a factor <strong>of</strong> cos 2 . This<br />
relationship is common in additional bistatic properties. For example bistatic doppler is reduced<br />
by the factor cos and when beta is small the bistatic RCS is equal to the monostatic RCS<br />
2<br />
reduced by the factor cos 2 .<br />
In general any measurement made with a basic resolution <strong>of</strong> M, will have a root mean square<br />
(RMS) error M given by:[24]<br />
M<br />
M ≃<br />
(2.13)<br />
2⋅SNR<br />
Thus <strong>for</strong> the range resolution, the range error is given by:[24]<br />
R ≃<br />
c<br />
2 B 2⋅SNR<br />
(2.14)<br />
Thus the range accuracy is limited by the pulse effective bandwidth and SNR.<br />
35
It is clear from the above that the resolution and its accuracy is improved by using shorter pulses,<br />
i.e. Large pulse bandwidths. However, there's a lower limit to pulse duration. Depending on<br />
various parameters a minimum signal energy is required in order to detect a target echo.<br />
The energy carried by the pulse is a product <strong>of</strong> transmitter power and pulse duration. There<strong>for</strong>e,<br />
cutting pulse duration in half, <strong>for</strong> instance, would necessitate doubling transmitter power in order<br />
to keep the signal's energy constant. The problem with that is that transmitter power cannot be<br />
increased at will because <strong>of</strong> cost and other constraints. Increasing the rate at which pulses are<br />
transmitted by the radar will increase the mean power radiated and thus the energy radiated.<br />
However, radars are designed so that a range counter starts at the transmission <strong>of</strong> the pulse, is<br />
read out when an echo is detected, and is reset upon transmission <strong>of</strong> the next pulse. Thus when<br />
pulses are transmitted so frequently that a pulse is transmitted be<strong>for</strong>e the previous pulse has<br />
completed the round trip to the target and back. The resulting effect is an uncertainty in<br />
determining the correct relationship between the transmitted and the echo pulses and the target<br />
range becomes ambiguous. As depicted in Figure 2.3.<br />
The range beyond which targets appear as second time around echoes is called the maximum<br />
unambiguous range [29], which is given by:<br />
R M<br />
u<br />
=<br />
c<br />
2⋅PRF<br />
(2.15)<br />
where:<br />
PRF is the Pulse Repetition Frequency and is the rate at which pulses are transmitted<br />
The corresponding bistatic unambiguous range is given by:<br />
R T<br />
R R<br />
u<br />
=<br />
c<br />
PRF<br />
(2.16)<br />
which lies on an isorange contour, <strong>of</strong> major axis length equal to equation (B.1). Further details<br />
on unambiguous bistatic PRF can be found in [34] [24] [17].<br />
Figure 2.3: A target echo detected after the transmission <strong>of</strong> the second pulse results in ambiguous<br />
ranges at B and C.<br />
36
2.6.6 Doppler Resolution<br />
In order to accurately measure the speed <strong>of</strong> the target the doppler shift will be used. This is the<br />
change in radio frequency <strong>of</strong> the signal, caused by the relative motions <strong>of</strong> the target and receiver.<br />
With the target in motion, the radar to target range as well as the phase <strong>of</strong> the radar signal are<br />
continuously changing.<br />
This change in phase with respect to time is related to the doppler angular frequency and can be<br />
further shown [29] that the doppler frequency shift is given by:<br />
where:<br />
f d<br />
= 2v r<br />
= 2v r f 0<br />
c<br />
(2.17)<br />
f d<br />
f 0<br />
v r<br />
is the doppler frequency shift<br />
is the transmitted frequency<br />
is the radial velocity <strong>of</strong> the target with respect to the radar<br />
The definition <strong>of</strong> the bistatic doppler frequency shift is different to the monostatic case, above.<br />
The change in radio frequency is a result <strong>of</strong> the rate <strong>of</strong> change <strong>of</strong> the total path length travelled<br />
by the signal. In the monostatic case this change in path length is equal to a change in the targets<br />
range. For a bistatic system the doppler shift is given by:<br />
f d<br />
= − 1 <br />
d R T<br />
R R<br />
<br />
dt<br />
(2.18)<br />
The expression indicates a decrease in doppler frequency <strong>for</strong> an increasing path length [24].<br />
The doppler frequency shift provides the means to distinguish between stationary and moving<br />
targets. However, is not a measure <strong>of</strong> the radial velocity <strong>of</strong> the target with respect to the radar as<br />
with the monostatic case.<br />
The velocity resolution, which is the ability to separate two targets separated in doppler is<br />
synonymous with the doppler resolution and can be expressed by:<br />
f d<br />
= 1<br />
PRI<br />
(2.19)<br />
The longer the duration <strong>of</strong> the signal, the finer the resolution will be and the more accurately the<br />
doppler frequency shift can be determined.<br />
From equation (2.19) above, the accuracy <strong>of</strong> the doppler resolution is given by:<br />
f d<br />
≃<br />
1<br />
PRI⋅ 2⋅SNR<br />
(2.20)<br />
37
Thus it is noted that the requirement <strong>for</strong> velocity resolution and accuracy implies long duration<br />
wave<strong>for</strong>ms whereas <strong>for</strong> range resolution and accuracy implies short wave<strong>for</strong>ms. This paradox<br />
must be considered during the signal processing and wave<strong>for</strong>m design.<br />
2.6.7 Ambiguity Function<br />
The radar system's resolution, detection, measurement accuracy and ambiguity are defined by the<br />
radar wave<strong>for</strong>m [29]. The effect <strong>of</strong> the wave<strong>for</strong>m on these parameters maybe determined upon<br />
examination <strong>of</strong> the receiver output, which is optimum when designed on the matched filter<br />
among others [29] [24] [22]. Should a target be present where it is expected to be, the receiver<br />
output will peak to a maximum, in the absence <strong>of</strong> noise.<br />
Thus the probability <strong>of</strong> detection is independent <strong>of</strong> the shape <strong>of</strong> the wave<strong>for</strong>m and depends on<br />
the ratio <strong>of</strong> the energy in the signal to the noise power per cycle <strong>of</strong> bandwidth. Range and<br />
doppler accuracies are additionally dependant on this ratio, as previously discussed, additionally<br />
are affected by the shape <strong>of</strong> the radar wave<strong>for</strong>m as are ambiguity and resolution.<br />
Ambiguity occurs when more than one choice <strong>for</strong> a particular parameter is available but only one<br />
choice is expected. With respect to the matched filter receiver output, ambiguities arise when<br />
peaks occur at values other than the expected value.<br />
The effect <strong>of</strong> the transmitted wave<strong>for</strong>m on the doppler and range can be produced upon<br />
extension <strong>of</strong> the matched filter or autocorrelation function <strong>of</strong> the received signal. A three<br />
dimensional plot <strong>of</strong> echo turn around time, doppler frequency and the output <strong>of</strong> the function<br />
results in the ambiguity diagram which graphically presents the accuracy and ambiguity af<strong>for</strong>ded<br />
by the transmitted radar wave<strong>for</strong>m. Additionally, to examine the wave<strong>for</strong>m's potential to resolve<br />
two targets <strong>of</strong> equal amplitude with different doppler and range [29].<br />
The ambiguity plot is determined by the ambiguity function and as is given by:<br />
∣ R R , f d ∣ 2 ∞<br />
= ∣ ∫ s<br />
−∞ t<br />
t s ∗ 2<br />
t<br />
t−R R<br />
exp[ j 2 f d<br />
t ]dt ∣<br />
(2.21)<br />
where:<br />
R R<br />
f d<br />
is the time delay (or range)<br />
is the signal doppler shift<br />
R R<br />
, f d<br />
is the ambiguity response at R R and f d<br />
The ideal ambiguity diagram, also know as the thumbtack surface, would consist <strong>of</strong> a single peak<br />
with a minuscule thickness at the origin and zero elsewhere [29]. This single spike prevents<br />
ambiguities and permits the frequency and echo delay time to be determined simultaneously to a<br />
high degree <strong>of</strong> accuracy. Furthermore, no matter their position, any two targets are resolvable.<br />
However, the fundamental properties <strong>of</strong> the ambiguity function prevent this type <strong>of</strong> behaviour.<br />
The main restrictions affecting the per<strong>for</strong>mance are that the maximum height <strong>of</strong> the function, and<br />
there<strong>for</strong>e the peak at the origin are <strong>of</strong> a fixed height. In addition to this, the volume enclosed by<br />
the function is fixed and finite.<br />
38
Typically, the function will be spread about the origin and the maximum peak height will be no<br />
greater than the fixed maximum amount thus the remaining energy is spread through the<br />
remainder <strong>of</strong> the diagram which may result in one or more additional peaks, perhaps has high as<br />
the one at the origin. The ambiguity diagram is centred at the origin, thus the response at<br />
R R<br />
, 0 is the response to reflections at a different range but at the same doppler as the<br />
target. 0, f d<br />
is the response to reflections at the same range as the target but at different<br />
doppler shifts.<br />
In PCL radar, there are two factors which affect the ambiguity function. These are the geometry<br />
<strong>of</strong> the system and the instantaneous modulation <strong>of</strong> the reference signal.<br />
The dependence <strong>of</strong> the ambiguity function <strong>of</strong> a bistatic radar on geometry has been evaluated by<br />
Tsao et al [40] [9] [17] [11]. They show that in the bistatic arrangement, time delay is no longer<br />
a linear function <strong>of</strong> target range and doppler is no longer a linear function <strong>of</strong> target velocity.<br />
Thus, the <strong>for</strong>m <strong>of</strong> the ambiguity function depends on the position and direction <strong>of</strong> motion <strong>of</strong> the<br />
target.<br />
Thus the bistatic ambiguity function <strong>of</strong> the signal<br />
st is given by:<br />
∣ R RH , R Ra , V H , V a , R , L ∣ 2 =<br />
(2.22)<br />
∣<br />
∞<br />
2<br />
−∞<br />
∫ s t<br />
t− a<br />
R Ra<br />
, R<br />
,Ls ∗ t<br />
t− H<br />
R RH<br />
, R<br />
, L<br />
exp [ j 2 f dH<br />
R RH<br />
,V H<br />
, R<br />
, L−2 f da<br />
R Ra<br />
, V a<br />
, R<br />
, Lt ]dt∣<br />
where:<br />
R RH<br />
R Ra<br />
V H<br />
V a<br />
f dH<br />
f da<br />
R<br />
L<br />
is the hypothesised range from the receiver to the target<br />
is the actual range from the receiver to the target<br />
is the hypothesised radial velocity <strong>of</strong> the target with respect to the receiver<br />
is the actual radial velocity <strong>of</strong> the target with respect to the receiver<br />
is the hypothesised doppler frequency<br />
is the actual doppler frequency<br />
is the angle from the receiver to the target with respect to North<br />
is the length <strong>of</strong> the baseline <strong>for</strong>med by the transmitter and receiver<br />
The reference point <strong>of</strong> the geometry is assumed is assumed to be the receiver. The important<br />
difference between this <strong>for</strong>m <strong>of</strong> the ambiguity function and the <strong>for</strong>m presented in Equation 2.21,<br />
is that the geometrical layout and relationship between the transmitter, receiver and target have<br />
been factored in. This can have a significant effect on the <strong>for</strong>m <strong>of</strong> the ambiguity function and the<br />
resulting range and doppler resolutions.<br />
39
It is important to know the properties <strong>of</strong> these ambiguity functions with time, as the variation in<br />
the <strong>for</strong>m <strong>of</strong> the ambiguity function would determine the radar’s per<strong>for</strong>mance.<br />
Thus investigation into the dependence <strong>of</strong> the ambiguity function on the instantaneous<br />
modulation or programme content <strong>of</strong> the particular signal being broadcast <strong>for</strong>ms, in part, the<br />
objective <strong>of</strong> this dissertation.<br />
2.7 Summary<br />
This chapter introduces and explores the concept <strong>of</strong> PCL and notes that fundamentally, PCL<br />
systems are bistatic in nature and thus measure the difference in time <strong>of</strong> arrival between the<br />
signal directly from the transmitter and the signal reflected <strong>of</strong>f the target. Furthermore, PCL<br />
systems make use non-co-operative transmitters using 'illuminators <strong>of</strong> opportunity' such as<br />
existing radar transmissions and broadcast and communication signals.<br />
Some <strong>of</strong> the unique advantages and disadvantages that arise as a result <strong>of</strong> the geometry and the<br />
use <strong>of</strong> illuminators <strong>of</strong> opportunity are listed. Although the per<strong>for</strong>mance <strong>of</strong> the bistatic<br />
arrangement does not rival that <strong>of</strong> the monostatic geometry, there are some distinct applications<br />
within the bistatic system which cannot be implemented by monostatic radars. An example<br />
whereby bistatic systems are a major advantage, is using such a system to detect stealthy aircraft<br />
and complete systems can be built on an extremely low budget.<br />
The history and progression <strong>of</strong> development and research in PCL systems is explored and thus<br />
highlighting the techniques developed by Howland, that express the viability with regards to the<br />
tracking accuracy, as well as the distance at which targets were detected. Furthermore, work<br />
presented by Baker and Griffiths examines the per<strong>for</strong>mance <strong>of</strong> a variety <strong>of</strong> potential wave<strong>for</strong>ms<br />
as the radar signal.<br />
Of the illuminators <strong>of</strong> opportunity available in the environment, broadcast transmitters present<br />
themselves as the most attractive <strong>for</strong> surveillance purposes, owing to their inherent properties <strong>of</strong><br />
high powers <strong>of</strong> transmission and widespread coverage.<br />
If such signals are to be used on the basis <strong>of</strong> PCL, it is necessary to know their behaviour in<br />
terms <strong>of</strong> their ambiguity function [36] i.e. resolution and sidelobe levels in range and doppler,<br />
and the effect <strong>of</strong> range and doppler ambiguities. These signal processing techniques are outlined<br />
at the end <strong>of</strong> the chapter, furthermore, the theory begins to highlight the purpose <strong>of</strong> work done in<br />
measuring the the radar's frequency reference to maintain timing accuracy.<br />
40
Chapter 3<br />
USRP and GNU<strong>Radio</strong><br />
3.1 Introduction<br />
The continuum <strong>of</strong> innovation in semiconductor technology has led to the development <strong>of</strong> new<br />
radio technologies. One such technology is s<strong>of</strong>tware defined radio. Thus far, no standard<br />
definition <strong>of</strong> s<strong>of</strong>tware radio has been generated. This can be largely attributed to the nature <strong>of</strong> the<br />
flexibility that s<strong>of</strong>tware defined radios <strong>of</strong>fer.<br />
As the interest and proliferation <strong>of</strong> this technology pushes ever widening frontiers, the Free<br />
S<strong>of</strong>tware Foundation (FSF) has developed and is maintaining a S<strong>of</strong>tware <strong>Defined</strong> <strong>Radio</strong> project<br />
called GNU<strong>Radio</strong>. This GNU project is a collection <strong>of</strong> s<strong>of</strong>tware tools that can be used to<br />
implement signal processing and radio applications on a PC using external USB based hardware,<br />
known as the Universal S<strong>of</strong>tware <strong>Radio</strong> Peripheral (USRP) also developed as part <strong>of</strong><br />
GNU<strong>Radio</strong>.<br />
This chapter presents the research into the SDR paradigm and the associated design<br />
considerations, thus providing the natural backdrop against which the per<strong>for</strong>mance <strong>of</strong> the USRP<br />
and GNU<strong>Radio</strong> are examined.<br />
3.2 S<strong>of</strong>tware <strong>Defined</strong> <strong>Radio</strong><br />
However, SDRs do have characteristics that make them unique in comparison to other types <strong>of</strong><br />
radios. SDRs have the ability to be trans<strong>for</strong>med through the use <strong>of</strong> s<strong>of</strong>tware and redefinable<br />
logic. This is achieved with general purpose digital signal processors (DSPs) or field<br />
programmable gate arrays (FPGAs). Thus SDRs have the ability to go beyond simple single<br />
channel, single mode transceiver technology with the ability to change modes arbitrarily because<br />
the channel bandwidth, rate, and modulation are all flexibly determined through s<strong>of</strong>tware. This is<br />
unlike most radios in which the processing is done with either analogue circuitry or analogue<br />
circuitry combined with digital chips.<br />
41
As an example, a European GSM SDR device may load a codec to process American GSM<br />
signals on detecting a change in the underlying service network, continuing to demodulate GSM<br />
signalling with no change in the physical hardware [20].<br />
In order to make full use <strong>of</strong> this concept, SDRs keep the signal in the digital domain <strong>for</strong> as much<br />
<strong>of</strong> the signal chain as possible, digitizing and reconstructing close to the antenna, thereby<br />
allowing digital techniques to per<strong>for</strong>m functions done by analogue components as well as the<br />
additional functions that are not possible in the analogue domain. To achieve this digitisation,<br />
ADCs, in the receive chain and DACs, in the transmit chain, are employed. However, connecting<br />
these components directly to the antenna introduces concerns regarding selectivity and<br />
sensitivity.<br />
Thus the need <strong>for</strong> a flexible analogue front end capable <strong>of</strong> translating a wide range <strong>of</strong><br />
frequencies and bands to that which the data converters are able to adequately process [25].<br />
Some <strong>of</strong> the typical dynamic characteristics <strong>of</strong> an SDR include:<br />
●<br />
●<br />
●<br />
●<br />
Channel Bandwidth<br />
Data rates<br />
Modulation type<br />
Conversion Gain<br />
This is exemplified in multiband radio systems, these systems are capable <strong>of</strong> operating on two or<br />
more frequency bands sequentially or simultaneously. Typically multiple radios would be<br />
required, each limited to operation in a specified band. In addition to this, multicarrier or<br />
multichannel radios have the ability to simultaneously operate on more than one frequency at a<br />
time, whether or not they fall in the same frequency band.<br />
Multimode systems process several different kinds <strong>of</strong> standards and can be continuously<br />
reprogrammed. Examples <strong>of</strong> such standards include AM, FM, GMSK, and CDMA. Furthermore,<br />
a traditional radio determines the channel bandwidth with a fixed analogue filter. However, a<br />
SDR determines the channel bandwidth using digital filters that can be altered. This provides<br />
SDRs with the resources to vary the bandwidth dynamically, additionally, digital filters have the<br />
potential to implement filters not possible in the analogue domain and are able compensate <strong>for</strong><br />
transmission path distortion [7]. These features are complex to implement using analogue filters.<br />
3.2.1 Applications <strong>of</strong> S<strong>of</strong>tware <strong>Radio</strong><br />
S<strong>of</strong>tware <strong>Radio</strong> is finding use in an ever broadening spectrum <strong>of</strong> applications. S<strong>of</strong>tware radio<br />
enable rapid prototyping and deployment <strong>of</strong> sophisticated wireless systems, making it ideal <strong>for</strong> a<br />
plethora <strong>of</strong> research and development applications in academia, industry and defence.<br />
To name just a few, s<strong>of</strong>tware radio systems have been deployed in:<br />
●<br />
●<br />
●<br />
●<br />
●<br />
Public Safety and mine/underground communications<br />
Battlefield, survivable and ad-hoc networks<br />
Passive Coherent Radar<br />
Cognitive <strong>Radio</strong><br />
<strong>Radio</strong> Astronomy<br />
42
3.2.2 S<strong>of</strong>tware <strong>Radio</strong> development Tools and Frameworks<br />
Several S<strong>of</strong>tware <strong>Radio</strong> environments are available. The common and favoured solutions<br />
include:<br />
●<br />
●<br />
●<br />
●<br />
●<br />
C/C++<br />
Matlab/Simulink/LabView<br />
JTRS<br />
OSSIE<br />
GNU<strong>Radio</strong><br />
This dissertation focuses its investigation on GNU<strong>Radio</strong>.<br />
3.3 Anatomy <strong>of</strong> a SDR Receiver<br />
The first element in the receiver chain is the antenna. In a fully developed SDR the antenna is<br />
additionally a programmable component. The antenna generally tends to be the weakest element<br />
<strong>of</strong> the receiver. This is primarily so as the antenna has a bandwidth that is a small percentage <strong>of</strong><br />
the centre frequency and there<strong>for</strong>e, multiband operation is difficult. However, in instances where<br />
single bands <strong>of</strong> operation are used, this is not a problem.<br />
In the ideal scenario, the SDR would have its ADC connected directly to the antenna <strong>of</strong> the<br />
receiver, this is in order to facilitate the implementation <strong>of</strong> as much <strong>of</strong> the system components in<br />
the digital domain. This, however, is not a practical solution and some <strong>for</strong>m <strong>of</strong> an analogue front<br />
end must be used be<strong>for</strong>e the ADC in the receive path in order to conduct the appropriate<br />
frequency translation.<br />
The most common <strong>of</strong> these architectures is the super heterodyne architecture. In this instance the<br />
midrange IF produced by the receiver allows the use <strong>of</strong> sharper cut <strong>of</strong>f filters <strong>for</strong> improved<br />
selectivity. Additionally higher IF gains are achieved through the use <strong>of</strong> IF amplifiers, improving<br />
the sensitivity. The super heterodyne receiver uses two stages <strong>of</strong> conversion and is useful at<br />
microwave frequencies in avoiding problems associated with Local Oscillator (LO) instability.<br />
The super heterodyne receiver <strong>of</strong>fers this per<strong>for</strong>mance over a range <strong>of</strong> frequencies, thus, by<br />
combining wideband analogue techniques and multiple front ends facilitates the operation <strong>of</strong> the<br />
receiver across different RF bands.<br />
The direct conversion architecture, which converts the RF signal directly to an IF frequency <strong>of</strong><br />
zero, in less demanding applications, is seeing an increase in popularity. This architecture has the<br />
advantage <strong>of</strong> being simpler to implement and less costly to build. This is so as they have no IF<br />
amplifier, IF bandpass filter or IF LO required <strong>for</strong> the final down-conversion. Additionally, the<br />
direct conversion approach does not generate any image frequencies as the mixer difference<br />
frequency is effectively zero, while the sum frequency is easily filtered out.<br />
However, this architecture has a major drawback in that the LO must have a very high degree <strong>of</strong><br />
precision and stability to avoid drifting <strong>of</strong> the received signal frequency. Currently, direct<br />
conversion is found in user terminals <strong>for</strong> cellular communication [23]. Additionally, direct<br />
conversion receivers are susceptible to various noise sources at DC, which create a DC <strong>of</strong>fset<br />
which may be larger than the signal itself and will degrade the SNR <strong>of</strong> the signal [1].<br />
43
Band select filters are made use <strong>of</strong> in order to limit the range <strong>of</strong> input frequencies presented to<br />
the high gain stage. This is in aid <strong>of</strong> minimising the effects <strong>of</strong> intermodulation distortion. In<br />
addition, where intermodulation is not a problem, it is still possible that strong out <strong>of</strong> band<br />
signals could limit the amount <strong>of</strong> potential gain in the following stages, resulting in limited<br />
sensitivity. If all the signals input are <strong>of</strong> interest and filtering <strong>of</strong> the stronger signals is not<br />
possible, it becomes necessary <strong>for</strong> the receiver to have a large dynamic range.<br />
Mixers are used to translate the RF spectrum to a suitable IF frequency. Receivers may use a<br />
number <strong>of</strong> mixer stages, each successively generating a lower frequency, as was described in the<br />
super heterodyne architecture. In order to eliminate undesired images as well as other undesired<br />
signals, filtering is made use <strong>of</strong> at each stage. The filtering should also be appropriate <strong>for</strong> the<br />
application. For example a traditional single carrier receiver would generally apply channel<br />
filtering through the mixer stages to help control the IP3 requirements <strong>of</strong> each stage [7].<br />
A quadrature demodulator in addition to, or instead <strong>of</strong>, a mixer is employed in some receiver<br />
architectures. This is in order to separate the I and Q components which undergo separate signal<br />
conditioning. Due to the digital nature <strong>of</strong> the receiver this is not a problem. However, in the<br />
analogue domain the signal paths must be perfectly matched, or I/Q imbalances will be<br />
introduced, potentially limiting the suitability <strong>of</strong> the system.<br />
An IF amplifier is <strong>of</strong>ten made use <strong>of</strong> in the <strong>for</strong>m <strong>of</strong> an Automatic Gain Control (AGC). The goal<br />
<strong>of</strong> the AGC is to use the maximum gain possible without saturating the the signal chain. If the<br />
dynamic range in the receiver, which is a function <strong>of</strong> the ADC per<strong>for</strong>mance, is insufficient,<br />
reduction in gain from a strong signal may cause weaker signals to be lost in the noise floor <strong>of</strong><br />
the receiver.<br />
The ADC is used to convert the IF signal or signals into digital <strong>for</strong>mat <strong>for</strong> processing. The ADC<br />
capabilities define the per<strong>for</strong>mance <strong>of</strong> the the SDR receiver. It is common place <strong>for</strong> the ADCs in<br />
SDR receivers to be over specified as their applications are unknown prior to their selection and<br />
the best available ADC is selected.<br />
A number <strong>of</strong> options are available to implement the digital preprocessor. For very high sample<br />
and data rates, this is usually implemented as a FPGA, which by nature are flexible in their<br />
functions and range <strong>of</strong> parameters. An FPGA can, <strong>of</strong> course, be programmed <strong>for</strong> any function<br />
desired. Typically, an FPGA would be programmed to per<strong>for</strong>m the quadrature demodulation and<br />
tuning, channel filtering, and data rate reduction among others. The FPGA is capable <strong>of</strong><br />
implementing and per<strong>for</strong>ming most signal processing tasks and control all <strong>of</strong> the features in the<br />
other elements.<br />
3.4 Universal S<strong>of</strong>tware <strong>Radio</strong> Peripheral<br />
The USRP is a hardware plat<strong>for</strong>m that encompasses the s<strong>of</strong>tware radio paradigm. It has been<br />
designed to interface primarily with the GNU<strong>Radio</strong> s<strong>of</strong>tware, thereby providing a cheap, flexible<br />
and high speed s<strong>of</strong>tware radio prototyping test bed. The USRP is comprised <strong>of</strong> the motherboard,<br />
which houses a FPGA <strong>for</strong> high speed signal processing, and interchangeable daughterboards<br />
that cover different frequency ranges. The average cost <strong>of</strong> the USRP motherboard and a single<br />
daughterboard is US$1000.00<br />
44
The USRP provides several functions; digitization <strong>of</strong> the input signal, digital tuning within the IF<br />
band, and sample rate reduction be<strong>for</strong>e sending the digitized baseband data to the computing<br />
plat<strong>for</strong>m via the USB interface. It provides the opposite processing functions <strong>for</strong> the transmit<br />
path.<br />
Figure 3.1 below shows the high level block diagram <strong>of</strong> the USRP board with daughterboards<br />
and Figure 3.2 shows an image <strong>of</strong> the USRP with the basic daughterboards plugged in.<br />
Figure 3.1: USRP and daughterboard block diagram [5]<br />
3.4.1 FX2 USB 2 Controller<br />
The FX2 microcontroller contains an embedded USB 2.0 transceiver and handles all USB<br />
transfers with the upstream USB host. It presents a data bus to the FPGA, with generic control<br />
signals which can be programmed to behave in a custom manner. This interface is called the<br />
General Purpose Interface (GPIF).<br />
The FX2 also handles all USB control requests, which all USB enabled devices must support to<br />
fully comply with the USB standard. These include responses to device capability interrogations<br />
and standard setup requests [32].<br />
3.4.2 ADCs and DACs<br />
The analogue interface portion houses the Analog Devices, AD9862. The AD9862 provides<br />
several functions. Each receive section contains two ADCs. The ADCs operate at 64<br />
Msamples/second with 12 bits <strong>of</strong> resolution and with reference to the Nyquist criterion the ADCs<br />
are capable <strong>of</strong> digitising a band 32MHz wide.<br />
Positioned ahead <strong>of</strong> the ADCs there are programmable gain amplifiers (PGA) with 20dB <strong>of</strong> gain<br />
available to adjust the input signal level in order to maximise use <strong>of</strong> the ADCs dynamic range<br />
which evaluates to approximately 72dB <strong>of</strong> dynamic range. The full range on the ADCs is 2V<br />
peak to peak, and the input is 200 ohms differential. This is equates to 5mW, or 7dBm.<br />
45
Figure 3.2: USRP and daughterboards [39]<br />
The transmit path provides an interpolater and upconverter to match the output sample rate to the<br />
DAC sample rate and convert the baseband input to a low IF output. There are PGAs after the<br />
DACs. The DACs operate at 128 Msamples/second with 14 bits <strong>of</strong> resolution and can supply a<br />
50 ohm differential load, with 10mW or 10dBm. There is also PGA used after the DAC,<br />
providing up to 20dB gain.<br />
This is shown in the ADC block diagram, in Figure 3.3.<br />
In any digitisation process, the faster that the signal is sampled, results in an effective gain <strong>of</strong><br />
received SNR.<br />
46
Furthermore, the relationship between the SNR and sampling rate in the signal bandwidth is<br />
given by:<br />
where:<br />
N<br />
f s<br />
B<br />
SNR = 6.02 N 1.76dB 10 log f s<br />
2B<br />
is the number <strong>of</strong> bits <strong>of</strong> resolution <strong>of</strong> the ADC<br />
is the sampling rate <strong>of</strong> the ADC<br />
is the Bandwidth <strong>of</strong> the signal<br />
(3.1)<br />
Figure 3.3: AD9862 ADC block Diagram [38]<br />
47
3.4.3 FPGA<br />
The FPGA employed by the USRP is the Altera Cyclone and per<strong>for</strong>ms the high speed signal<br />
processing. Furthermore, the FPGA manages the signal including reducing the data rates to<br />
facilitate their transport over the USB2 interface chip, the Cypress FX2, to the host PC where<br />
less intensive processing takes place. The FPGA and the FX2 chip are programmed over the<br />
USB2 bus.<br />
The clock provided by the USRP drives the ADCs at 64 Msps. If needed the AD9862 may divide<br />
this clock by two to reduce the sample rate. This only affects the clock rate <strong>of</strong> the ADCs, most <strong>of</strong><br />
the sample rate conversion is done in the FPGA and is referred to as decimation in this case.<br />
The block diagram <strong>of</strong> the USRP board and the high level FPGA signal processing blocks is<br />
shown in the image below.<br />
Figure 3.4: USRP and FPGA block Diagram [39]<br />
Following the signals digitisation, the data is sent to the FPGA. The standard FPGA firmware<br />
provides two Digital Downconverters (DDC), a second FPGA implementation however provides<br />
an additional two DDCs. The FPGA uses a multiplexer to connect the input streams from each <strong>of</strong><br />
the ADCs to the inputs <strong>of</strong> the DDCs.<br />
48
The multiplexer allows the USRP to support both real and complex input signals and allows <strong>for</strong><br />
having multiple channels selected out <strong>of</strong> the same ADC sample stream. The DDCs operate as<br />
real downconverters using the data from one ADC fed into the real channel or as complex DDCs<br />
where the data from one ADC is fed to the real channel and the data from another ADC is fed to<br />
the complex channel via the multiplexer.<br />
The DDC consists <strong>of</strong> a numerically controlled oscillator (NCO), a digital mixer, and a cascaded<br />
integrator comb (CIC) filter. These components downconvert the desired channel from the IF to<br />
baseband, reduce the sample rate and provide low pass filtering. The effect <strong>of</strong> the NCO and<br />
multiplier are implemented using the CORDIC algorithm.<br />
This shown in Figure 3.5.<br />
Figure 3.5: DDC block Diagram [39]<br />
Since the USB bus operates at a maximum rate <strong>of</strong> 480 Mbps the FPGA must reduce the sample<br />
rate in the receive path. The decimator can be treated as a low pass filter followed by a<br />
downsampler. Suppose the decimation factor is D. If we look at the digital spectrum, the low<br />
pass filter selects out the band [−/ D , / D] and then the downsampler spreads the spectrum<br />
to [− ,] . So in fact, we have narrowed the bandwidth <strong>of</strong> the digital signal <strong>of</strong> interest by a<br />
factor <strong>of</strong> D.<br />
The transmit path <strong>for</strong> the USRP is similar to the receive path, however there are differences.<br />
Since the sample rate the DACs operate at 128 Msps, it is necessary to increase the sample rate<br />
in order to match the sample rates between the high speed data converter and the lower speeds<br />
supported by the USB connection, this process is called interpolation and an interpolater running<br />
on the FPGA increases the data rate. The AD9862 also provides a further data rate increase by a<br />
factor <strong>of</strong> four. The transmit portion <strong>of</strong> the AD9862 provides the mixer and NCO required to set<br />
the IF frequency <strong>of</strong> the transmitted signal, the FPGA per<strong>for</strong>ms this function in the receive path.<br />
This is shown in Figure 3.6.<br />
49
Figure 3.6: USRP transmit chain block Diagram [39]<br />
3.4.4 The Daughterboards<br />
On the motherboard there are four slots into which you can plug in up to 2 RX daughterboards<br />
and 2 TX daughterboards. The daughterboards are used to hold the the RF receiver interface or<br />
tuner and the RF transmitter.<br />
There are slots <strong>for</strong> 2 TX daughterboards, labelled TXA and TXB, and 2 corresponding RX<br />
daughterboards, RXA and RXB. Each daughterboard slot has access to 2 <strong>of</strong> the 4 high-speed AD<br />
/DA converters (DAC outputs <strong>for</strong> TX, ADC inputs <strong>for</strong> RX). This allows each daughterboard<br />
which uses real (not IQ) sampling to have 2 independent RF sections, and 2 antennas (4 total <strong>for</strong><br />
the system). If complex IQ sampling is used, each board can support a single RF section, <strong>for</strong> a<br />
total <strong>of</strong> 2 <strong>for</strong> the whole system.<br />
A variety <strong>of</strong> daughterboards are available, including [6]:<br />
Basic Tx and Rx Boards<br />
The BasicTX and BasicRX are designed <strong>for</strong> use with external RF frontends as an IF interface.<br />
The ADC inputs and DAC outputs are directly trans<strong>for</strong>mer-coupled to SMA connectors 50 Ohm<br />
impedance with no mixers, filters, or amplifiers.<br />
The BasicTX and BasicRX operate in the 1MHz to 250MHz range and give direct access to all<br />
<strong>of</strong> the signals on the daughterboard interface including 16 bits <strong>of</strong> high-speed digital I/O, SPI and<br />
I2C buses, and the low-speed ADCs and DACs, and as such are useful <strong>for</strong> developing custom<br />
daughterboards or FPGA designs.<br />
LFTX and LFRX Boards<br />
The LFTX and LFRX are very similar to the BasicTX and BasicRX, respectively, with 2 main<br />
differences. Because the LFTX and LFRX use differential amplifiers instead <strong>of</strong> trans<strong>for</strong>mers,<br />
their frequency response extends down to DC. The LFTX and LFRX also have 30 MHz low<br />
pass filters <strong>for</strong> anti aliasing.<br />
50
DBSRX<br />
The DBSRX is a complete receiver system <strong>for</strong> 800 MHz to 2.4 GHz with a 3-5 dB noise figure.<br />
The DBSRX features a s<strong>of</strong>tware controllable channel filter which can be made as narrow as<br />
1 MHz, or as wide as 60 MHz.<br />
The DBSRX frequency range covers many bands <strong>of</strong> interest, including all GPS and Galileo<br />
bands, the 902-928 MHz ISM band, cellular and PCS, the Hydrogen and Hydroxyl radio<br />
astronomy bands, DECT, and others. The DBSRX is MIMO capable, and can power an active<br />
antenna via the coax.<br />
RFX<br />
The RFX family <strong>of</strong> daughterboards turns a USRP motherboard into a complete RF transceiver<br />
system. An antenna is necessary in order to achieve two-way, high bandwidth communications<br />
in many popular frequency bands. The family <strong>of</strong> RFX daughterboards cover a frequency range<br />
from 400MHz to 2.9GHz. The boards have many features which facilitate their integration into<br />
more complex systems, such as digital control lines and the option <strong>for</strong> split transmit and receive<br />
ports.<br />
TVRX<br />
The TVRX daughterboard is a complete VHF and UHF receiver system based on a TV tuner<br />
module. It operates over a 6 MHz wide block <strong>of</strong> spectrum from anywhere in the 50-860 MHz<br />
range. All tuning and AGC functions can be controlled from s<strong>of</strong>tware.<br />
Different tuner modules have been used <strong>for</strong> the TVRx. They differ in the frequency <strong>of</strong> the IF<br />
output and whether or not they per<strong>for</strong>m a single-conversion or a second down-conversion. In this<br />
dissertation the module using a single-conversion approach to 43.75MHz has been used. The<br />
alternative module <strong>of</strong>fers a second conversion to 5.75MHz.<br />
This board will be used as the front end <strong>for</strong> the PCL system and thus is used in all the receiver<br />
tests conducted and <strong>for</strong>ms the subject <strong>of</strong> the receiver characterisation presented in this<br />
dissertation.<br />
Figure 3.7: Signal flow from a single TVRx daughterboard to DDC0<br />
51
The block diagram in Figure 3.7 is representative <strong>of</strong> the flow path <strong>of</strong> the TVRx signal. The<br />
daughterboard produces a differential pair as output, shown as the signal S(t) and -S(t). These<br />
signals, after they have been digitised, are routed via the multiplexer to the inputs <strong>of</strong> the DDC0.<br />
It is at this stage that the complex signal I(n) + jQ(n) is produced. In the instances where the<br />
second TVRx daughterboard is used, a second differential pair is produced, digitised at ADC2<br />
and ADC3 respectively and routed to DDC1 through the multiplexer. Thus two IQ pairs are sent<br />
from the USRP to the PC via the USB <strong>for</strong> further processing.<br />
3.5 GNU<strong>Radio</strong><br />
GNU<strong>Radio</strong> is an open source, s<strong>of</strong>tware defined, radio plat<strong>for</strong>m <strong>for</strong> building and deploying<br />
s<strong>of</strong>tware radios. It has a large worldwide community <strong>of</strong> developers and users that have<br />
contributed to a substantial code base and provided many practical applications <strong>for</strong> the hardware<br />
and s<strong>of</strong>tware.<br />
It provides a complete development environment to create s<strong>of</strong>tware radios, handling all <strong>of</strong> the<br />
hardware interfacing, multi threading and portability issues. GNU <strong>Radio</strong> has libraries <strong>for</strong> all<br />
common s<strong>of</strong>tware radio needs, including;<br />
various modulations:<br />
●<br />
●<br />
●<br />
GMSK<br />
PSK<br />
OFDM<br />
error-correcting codes:<br />
●<br />
●<br />
●<br />
Reed-Solomon<br />
Viterbi<br />
Turbo Codes<br />
signal processing constructs<br />
●<br />
●<br />
●<br />
●<br />
●<br />
optimized filters<br />
FFTs<br />
equalizers<br />
timing recovery<br />
Scheduling<br />
It is a very flexible system, it is user expandable and it allows applications to be developed in<br />
C++ or Python.<br />
3.5.1 The GNU<strong>Radio</strong> Architecture<br />
GNU <strong>Radio</strong> is organised using a two-tier structure that provides a data flow abstraction. It builds<br />
applications using python code <strong>for</strong> high level organisation, policy, GUI and other non<br />
per<strong>for</strong>mance critical functions, by employing the concept <strong>of</strong> a graph containing signal processing<br />
blocks and connections <strong>for</strong> data flow between them.<br />
52
The signal processing blocks are implemented and correspond to some sophisticated functions or<br />
class methods in C++.<br />
Many processing blocks currently exist within the GNU<strong>Radio</strong> framework and include filters,<br />
demodulators and other signal manipulation elements.<br />
A graph can be thought <strong>of</strong> as a framework upon which all the necessary elements are placed and<br />
then connected. Using a schematic diagram as an analogy <strong>for</strong> a graph, the elements placed on the<br />
schematic include the resistors, amplifiers, etc. These components are then “wired” up <strong>for</strong><br />
current to flow between them. Thus in GNU <strong>Radio</strong>, the circuit parts are replaced by signal<br />
sources, sinks and the signal processing blocks.<br />
Figure 3.8: GNU<strong>Radio</strong> application structure<br />
Conceptually, blocks process infinite streams <strong>of</strong> data flowing from their input ports to their<br />
output ports. Blocks’ attributes include the number <strong>of</strong> input and output ports they have as well as<br />
the type <strong>of</strong> data that flows through each. The most frequently used types are short, float and<br />
complex. The Simplified Wrapper and Interface Generator (SWIG), is what enables the C++<br />
implementation <strong>of</strong> the various blocks to be used from Python.<br />
Some blocks have only output ports or input ports. These serve as data sources and sinks in the<br />
graph. The signal source can have various implementations, but is usually the USRP or a file<br />
containing sampled data. The signal stream then flows through any number <strong>of</strong> processing nodes.<br />
The final output <strong>of</strong> the processing graph terminates in a signal sink, which may be a real time<br />
spectrum analyser, a real time oscilloscope, an output file, the USRP or audio hardware. It is<br />
estimated that 100 blocks are implemented in GNU<strong>Radio</strong> and new blocks can be written as<br />
desired.<br />
This framework allows <strong>for</strong> simple implementation <strong>of</strong> powerful signal processing systems.<br />
53
An elementary application is presented here to present the concepts.<br />
#>>> bring in blocks from the main gnu radio package<br />
from gnuradio import gr<br />
#>>> bring in the audio source/sink<br />
from gnuradio import audio<br />
#>>> create the flow graph<br />
fg = gr.flow_graph()<br />
#>>> create the signal sources<br />
#>>> parameters: samp_rate, type, output freq, amplitude, <strong>of</strong>fset<br />
src = gr.sig_source_f(32000, gr.GR_SIN_WAVE, 350, .5, 0)<br />
#>>> create a signal sink<br />
#>>> parameters: samp_rate<br />
sink = audio.sink(32000)<br />
#>>> connect the source to the sink<br />
fg.connect(src, sink)<br />
#>>> run the flow graph<br />
fg.run()<br />
3.5.2 Common GNU<strong>Radio</strong> Blocks<br />
As mentioned earlier, an estimated 100 signal processing blocks are predefined in the GNU<br />
<strong>Radio</strong> environment. The best way to become acquainted with these is to explore the GNU <strong>Radio</strong><br />
Documentation. This can be found on line at http://gnuradio.org/doc/doxygen/index.html.<br />
For demonstration purposes, two <strong>of</strong> the frequently used blocks will be mentioned here.<br />
Block: usrp.source_c [s]<br />
Usage:<br />
usrp.source_c (s) (int which_board,<br />
unsigned int decim_rate,<br />
int nchan = 1,<br />
int mux = -1,<br />
int mode = 0 )<br />
The suffix c (complex), or s (short) specifies the data type <strong>of</strong> the stream from USRP. Most likely<br />
the complex source (I/Q quadrature from the digital down converter (DDC)) would be more<br />
frequently used. which_board specifies which USRP to open, which is probably 0 if there is only<br />
one USRP board. decim_rate tells the digital down converter (DDC) the decimation factor D.<br />
nchan specifies the number <strong>of</strong> channels, 1, 2 or 4. mux sets input MUX configuration, which<br />
determines which ADC (or constant zero) is connected to each DDC input ‘-1’ means we<br />
preserve the default settings. mode sets the FPGA mode, which we seldom touch.<br />
54
The default value is 0, representing the normal mode. Quite <strong>of</strong>ten we only specify the first two<br />
arguments, using the default values <strong>for</strong> the others.<br />
For example:<br />
usrp_decim = 250<br />
src = usrp.source_c (0, usrp_decim)<br />
Block: calc_dxc_freq ()<br />
Usage:<br />
usrp.calc_dxc_freq(number target_freq,<br />
number baseband_freq,<br />
number fs)<br />
This function is a utility to calculate the frequency to be used <strong>for</strong> setting the digital up or down<br />
converter. It returns a 2-tuple (ddc_freq, inverted) where ddc_freq is the value <strong>for</strong> the DDC and<br />
inverted is True if we are operating in an inverted Nyquist zone. target_freq specifies the desired<br />
RF frequency in Hz. baseband_freq is the RF frequency that corresponds to DC in the IF and fs<br />
is the converter sample rate.<br />
3.5.3 Hardware<br />
Various hardware options are available <strong>for</strong> use with GNU <strong>Radio</strong>. For instance, the sound card<br />
can be used as an output sink. Wide band I/O and VXI/cPCI cards can also be used as additional<br />
options. However, largely identified as the primary hardware choice as regards GNU<strong>Radio</strong> is the<br />
USRP.<br />
3.6 Summary<br />
Included in the objectives <strong>of</strong> the overall PCL project is the development <strong>of</strong> a radar system at<br />
considerably low cost. In pursuit <strong>of</strong> this, SDR has been identified as a candidate technology. In<br />
this chapter a presentation <strong>of</strong> the definition and paradigm <strong>of</strong> SDR is made, including their<br />
applications and the development tools available.<br />
Furthermore, the considerations in the design <strong>of</strong> SDR receivers is made, facilitating the<br />
characterisation and testing <strong>of</strong> the SDR toolkit selected <strong>for</strong> the implementation <strong>of</strong> our radar<br />
receiver.<br />
In the light <strong>of</strong> this, the USRP and GNU<strong>Radio</strong> are examined, the USRP provides a flexible and<br />
powerful FPGA, developed by Altera, as the processing unit, which implements a great deal <strong>of</strong><br />
the signal processing in the receive and transmit chains. Furthermore, the USRP hosts two<br />
Analog Devices AD9862 chips, each providing a two ADCs sampling at a rate <strong>of</strong> 64<br />
Msamples/second with 12 bits <strong>of</strong> resolution and two DACs operating at 128 Msamples/second<br />
with 14 bits <strong>of</strong> resolution.<br />
55
Moreover, the AD9862 chips provide two auxiliary ADCs and three auxiliary DACs which can<br />
provide further flexibility in computation and control <strong>of</strong> external daughterboard components.<br />
These daughterboard boards provide front end flexibility allowing <strong>for</strong> the down-conversion and<br />
thus manipulation <strong>of</strong> the radio spectrum.<br />
GNU<strong>Radio</strong>, is seen to be a powerful toolkit that implements numerous complex signal<br />
processing elements and additionally interfaces with the USRP. Thus combining to provide a<br />
powerful and widely versatile toolset <strong>for</strong> the development <strong>of</strong> a variety <strong>of</strong> radio devices.<br />
The following chapter tests the USRP and GNU<strong>Radio</strong> in a variety <strong>of</strong> per<strong>for</strong>mance related issues<br />
and there<strong>for</strong>e, allowing <strong>for</strong> comment on its usefulness <strong>for</strong> our application.<br />
56
Chapter 4<br />
Receiver Characterisation<br />
4.1 Introduction<br />
The most critical component <strong>of</strong> a wireless system <strong>of</strong>ten is its receiver. The purpose <strong>of</strong> this is to<br />
extract the desired signal reliably from the various sources <strong>of</strong> signals, interference and noise.<br />
The following sections will present the characterisation <strong>of</strong> the USRP with the TVRx<br />
daughterboard and GNU<strong>Radio</strong> as a receiver in the PCL system being investigated.<br />
At this time, various aspects <strong>of</strong> the overall project are still under investigation and research. As a<br />
result, the specific requirements that the receiver system must meet are not available. However,<br />
the per<strong>for</strong>mance <strong>of</strong> GNU<strong>Radio</strong> in its control <strong>of</strong> the USRP with the TVRx daughterboard and<br />
some <strong>of</strong> the fundamental principles <strong>of</strong> radio receiver and digital radio receiver design are<br />
identified, discussed and measured, thereby facilitating this examination.<br />
The results <strong>of</strong> the characterisation will be presented within the body <strong>of</strong> each section, in order to<br />
facilitate coherency in its reading.<br />
4.2 Test <strong>System</strong> Overview<br />
This section will describe the manner in which the equipment was setup in order to measure the<br />
various characteristics described in this chapter.<br />
The test signal is generated by a Hewlett Packard 8656B signal generator, with the output<br />
frequency set to 100MHz. The signal generator has the desirable functionality to vary the power<br />
<strong>of</strong> the signal output generated. The signal is fed into the USRP, setup with the TVRx<br />
daughterboard, and after digitisation is sent to the PC. In addition to this, a Hewlett Packard<br />
54645D oscilloscope is used to probe the pin outs <strong>of</strong> the TVRx daughterboard. The individual<br />
test conditions will be further described in the relevant subsection.<br />
The test equipment is pictured in Figure 4.1.<br />
57
Figure 4.1: Experimental setup used in characterisation tests<br />
4.3 Receiver Bandwidth<br />
The TVRx daughterboard is built on the Microtune 4937 D15 RF tuner module. The<br />
instantaneous bandwidth by definition is the frequency band over which multiple signals are able<br />
to be simultaneously amplified to within a specified gain tolerance [29]. The bandwidth is an<br />
important property as it affects the noise and signal and has a direct impact on the level <strong>of</strong> signal<br />
processing gain achievable by the receiver as eluded to by Equation 2.6. This module provides a<br />
static 6MHz bandwidth, referenced to the centre frequency that the module is tuned to. With<br />
respect to our application, the signals <strong>of</strong> interest are the Broadcast FM signals and the target<br />
returns from the surveillance areas which have a bandwidth <strong>of</strong> 100kHz. Thus the receiver<br />
provides much more than the requisite bandwidth. Furthermore, we can sustain 32MB/second<br />
across the USB. All samples sent over the USB interface are in 16 bit signed integers in IQ<br />
<strong>for</strong>mat, i.e. 16 bit I data followed by 16 bit Q data, resulting in 8M complex samples/second<br />
across the USB. This provides a maximum effective total spectral bandwidth <strong>of</strong> about 8MHz.<br />
The FPGA provides the functionality <strong>of</strong> altering the decimation rate and thus allowing us to<br />
select narrower ranges. For instance, the bandwidth <strong>of</strong> a FM station is 100kHz, by selecting the<br />
decimation factor to be 250, the effective bandwidth across the USB is 64MHz / 250 = 256kHz,<br />
which is well suited <strong>for</strong> the 100kHz bandwidth without losing any spectral in<strong>for</strong>mation.<br />
58
4.4 Gain Range and Control<br />
The TVRx daughterboard has two AGC inputs available to level the signal. A block diagram <strong>of</strong><br />
this is presented in Figure 4.2. The daughterboard has a RF AGC range <strong>of</strong> 50dB and an IF AGC<br />
range <strong>of</strong> 33dB. GNU<strong>Radio</strong> achieves this gain variation by varying the voltage on to each <strong>of</strong> these<br />
inputs. As mentioned in the previous chapter the ADC that follows the daughterboard provides<br />
an additional level <strong>of</strong> amplification using its internal PGA, that provides an additional 20dB <strong>of</strong><br />
gain.<br />
Figure 4.2: Block diagram showing TVRx gain stages<br />
All three stages are set altogether via the set_gain(gain in dB) method. However, since multiple<br />
signals are to be processed simultaneously, the use <strong>of</strong> an AGC can result in a reduction <strong>of</strong> RF<br />
sensitivity and will cause weaker signals to be dropped. GNU<strong>Radio</strong> proves its versatility in this<br />
instance as the set_gain method can be bypassed and each gain stage set manually and<br />
independently.<br />
In the following tests conducted on the receiver and in those <strong>of</strong> later sections, the AGC was<br />
disabled and the amplifier stages were set independently. Figure 4.3 and Figure 4.4 show the<br />
variation <strong>of</strong> the voltage applied to the AGC pins <strong>of</strong> the daughterboards to the gain <strong>of</strong> the<br />
individual stages.<br />
The RF stage exhibits a linear increase in the gain, which accurately describes the behaviour <strong>of</strong><br />
the amplifier between the 10dB and 40dB range. Above 40dB the amplifier begins to run into<br />
saturation [37] and does not continue to show the linear conversion gain. The IF stage<br />
additionally exhibits a linear increase in gain. This is an accurate description <strong>of</strong> the amplifier<br />
behaviour in the range from 0dB to 15dB. Thus, GNU<strong>Radio</strong> provides us with an accurate<br />
description <strong>of</strong> the TVRx daughterboard's amplifier behaviour in the region from 10dB to 55dB.<br />
The PGA additionally, has a linear gain curve [38] and provides a further 20dB <strong>of</strong> gain,<br />
extending the gain range up to 75dB.<br />
59
Figure 4.3: Variation <strong>of</strong> voltage applied to RF AGC against RF Gain<br />
Figure 4.4: Variation <strong>of</strong> voltage applied to IF AGC against IF Gain<br />
60
4.5 Multiplexer Usage and Data Interleaving<br />
The FPGA multiplexer is like a router. It determines which ADC is connected to each DDC<br />
input. There are 4 DDCs and each has two inputs. GNU<strong>Radio</strong> controls the multiplexer using the<br />
usrp.set_mux() method.<br />
The multiple RX channels, which can be either 1,2, or 4, must all be the same data rate.<br />
There<strong>for</strong>e, the same decimation ratio must apply to all the channels. This condition additionally<br />
applies to the TX channels, however they may be different to the RX rate. When there are<br />
multiple channels, GNU<strong>Radio</strong> interleaves the channels. This means that with 4 channels, <strong>for</strong><br />
instance, the sequence sent over the USB would be:<br />
←<br />
DDC0 DDC1 DDC2 DDC3 DDC0 DDC1 DDC2<br />
I 0 Q 0 I 1 Q 1 I 2 Q 2 I 3 Q 3 I 0 Q 0 I 1 Q 1 I 2 Q 2<br />
←<br />
The USRP can operate in full duplex mode, the transmit and receive sides are completely<br />
independent <strong>of</strong> one another. The only consideration is that the combined data rate over the USB<br />
must be 32 MB/sec or less.<br />
The set_mux parameter is 32 bits, each nybble <strong>of</strong> 4 bits controls which ADC is connected to<br />
which DDC input. The least significant nybble <strong>of</strong> the parameter represents DDC0 and the most<br />
significant nybble <strong>of</strong> the parameter represents DDC3.<br />
DDC3 DDC2 DDC1 DDC0<br />
Q 3 I 3 Q 2 I 2 Q 1 I 1 Q 0 I 0<br />
Figure 4.5: Multiplexer setup <strong>for</strong> two TVRx to DDC0 and DDC1<br />
61
In Figure 4.5, the multiplexer parameter is set as Ox----3210, we are not concerned about the<br />
upper nybble as those bits set DDC3 and 2, represented here by dashes. However, we see that the<br />
I and Q values <strong>of</strong> DDC0 and DDC1 are set up accordingly.<br />
4.6 Frequency Range and Control<br />
The TVRx daughterboard that we are using uses a single conversion approach to 43.75MHz with<br />
the reception frequency divided into the VHF low, VHF high and UHF bands from 50MHz to<br />
860MHz. The band selection and tuning is done via the I 2 C bus, this coordinated from within<br />
GNU<strong>Radio</strong>. The tuning range is the frequency band over which the receiver will operate without<br />
degrading the specified per<strong>for</strong>mance [29].<br />
The oscillator frequency is driven by a 4MHz crystal reference and determined by the following:<br />
f osc<br />
= f ref<br />
× 8 × SF<br />
(4.1)<br />
where:<br />
f ref<br />
SF<br />
is the crystal reference frequency / Reference divider<br />
is the programmable scaling factor<br />
The reference divider mentioned above can take on one <strong>of</strong> three values, those being 512, 640 and<br />
1024. This value is set from within GNU<strong>Radio</strong> and the default is set at 640. The value <strong>of</strong> this<br />
reference divider influences the tuning step size, otherwise understood as the granularity <strong>of</strong> the<br />
tuner in setting the local oscillator frequency. The default choice results in a tuning step size <strong>of</strong><br />
50.0kHz with a potential minimum <strong>of</strong> 31.25kHz. This level <strong>of</strong> resolution is adequate <strong>for</strong> our<br />
application.<br />
GNU<strong>Radio</strong> sets up the appropriate byte values in the I 2 C data by use <strong>of</strong> the following method.<br />
Block: usrp.tune ()<br />
Usage:<br />
usrp.source_x.tune (u,<br />
chan,<br />
subdev,<br />
target_freq)<br />
Tuning is a two step process. First we ask the frontend to tune as close to the desired frequency<br />
as it can. Then we use the result <strong>of</strong> that operation and our target_frequency to determine the<br />
value <strong>for</strong> the digital down converter. The function returns False if there is a failure, in the case <strong>of</strong><br />
success it returns tune_result. U is the instance <strong>of</strong> the usrp that we want to tune. chan is the<br />
DDC channel that we wish to tune. Subdev tells the method which daughterboard subdevice to<br />
tune and target_freq is the centre frequency in Hz.<br />
62
However, GNU<strong>Radio</strong> does <strong>of</strong>fer the flexibility to control individual parts <strong>of</strong> the tuning process,<br />
e.g. tuning the TVRx frontend and setting the value <strong>of</strong> the digital down converter independently.<br />
4.7 Spurious Artifact<br />
During testing it was found that when a GNU<strong>Radio</strong> application is run, the USRP generates a<br />
large spurious output, this applies additionally to instances where the application is stopped and<br />
started again. This spurious output does corrupt data that is being stored to memory on the PC<br />
<strong>for</strong> <strong>of</strong>fline post processing. However, when the USRP is used <strong>for</strong> real time processing it does not<br />
present a concern, as the spurious output occurs at the start <strong>of</strong> the data. Figure 4.6 shows this<br />
artifact at the beginning <strong>of</strong> a data.<br />
Figure 4.6: Spurious signal generated by USRP on startup<br />
Figure 4.7 shows a zoomed in image <strong>of</strong> this artifact, illustrating the corruption <strong>of</strong> the first<br />
approximately 80 samples. In order to mitigate the effects <strong>of</strong> this artifact, this data is simply cut<br />
<strong>of</strong>f and the number <strong>of</strong> remaining samples made note <strong>of</strong> in the processing that follows.<br />
63
Figure 4.7: Closer look at the spurious signal generated by USRP on startup<br />
4.8 Receiver Noise Figure<br />
The noise figure is a measure <strong>of</strong> how much noise a component or system adds to a signal. The<br />
noise figure <strong>of</strong> a system depends on the losses in the circuit, the nature <strong>of</strong> the solid state devices,<br />
any bias applied and on amplification, to name but a few. When noise and a desired signal are<br />
applied to the input <strong>of</strong> a noiseless system, it is expected then that the noise and signal would be<br />
amplified or attenuated by the same factor, thus the SNR through the system would remain<br />
unchanged.<br />
A noisy system, however, will introduce noise. Thus the noise power will increase relative to the<br />
signal power, thereby reducing the SNR at the output <strong>of</strong> the system.<br />
The noise figure <strong>of</strong> a component is, by definition, independent <strong>of</strong> the noise input into the<br />
component and thus is based on a standard input noise source N i , at room temperature in a<br />
bandwidth B , thus:<br />
N i<br />
= T 0<br />
B<br />
(4.2)<br />
Where:<br />
<br />
T 0<br />
is the Boltzmann constant<br />
is the room temperature 290K<br />
64
Assuming a 1Hz bandwidth this value evaluates to the standard, 4x10 -21 W, which can be further<br />
expressed as -204dBW, which is further still expressed as -174dBm. The noise figure in essence<br />
is the difference between the noise output <strong>of</strong> the receiver and the noise output <strong>of</strong> an ideal,<br />
noiseless receiver.<br />
Noise figure is the decibel notation <strong>of</strong> the noise factor, which is defined as:<br />
F =<br />
SNR at input<br />
SNR at output = S i /N i<br />
S o<br />
/N o<br />
(4.3)<br />
In a typical receiver, the input signal will pass through a number <strong>of</strong> cascaded components such<br />
as filters, amplifiers and mixers, among others. Each stage will inject a degree <strong>of</strong> noise into the<br />
signal, degrading the signal to noise ratio. This makes it important to quantify the overall effect<br />
<strong>of</strong> the noise figure on the system's per<strong>for</strong>mance.<br />
Consider the network shown in the figure below, the total noise factor referenced to the input to<br />
the system is determined using the Friis equation as:<br />
F total<br />
= F 1<br />
F 2 − 1<br />
F 3 − 1<br />
F N<br />
− 1<br />
⋯ <br />
G 1<br />
G 1<br />
G 2<br />
G 1<br />
G 2<br />
⋯G N − 1<br />
(4.4)<br />
Figure 4.8: Cascaded components<br />
The characteristics <strong>of</strong> the cascaded system are dominated by the earlier stages. This is as the later<br />
stages are reduced by the product <strong>of</strong> the gains <strong>of</strong> the preceding stages. Thus the ideal scenario is<br />
one in which the first stage or component <strong>of</strong> a system has a low noise figure and a high gain to<br />
ensure the low noise figure <strong>for</strong> the overall system.<br />
In order to determine the noise figure <strong>of</strong> the receiver system, a known signal <strong>of</strong> low magnitude is<br />
input. In this instance -50dBm was input. The USRP was setup with a decimation rate <strong>of</strong> 64, thus<br />
providing an effective 1MHz <strong>of</strong> bandwidth <strong>for</strong> analysis. A spectrum analyser was setup, with<br />
1024 points <strong>of</strong> measurement resolution thus providing a reading representing 1kHz per<br />
measurement bin. This setup allows <strong>for</strong> the measurement <strong>of</strong> the noise floor per bin, by observing<br />
how far down the noise is from the input signal. The noise floor is a further 38dB below this<br />
level, measured in dBm/Hz. The noise figure is then determined by the difference between this<br />
signal and the theoretical level <strong>of</strong> -174dBm/Hz, derived from Equation 4.2. The noise figure <strong>of</strong><br />
the TVRx and the USRP is thus determined to be 10dB.<br />
65
4.9 Sensitivity<br />
The sensitivity <strong>of</strong> a receiver is defined as the lowest power level at which the receiver can detect<br />
an RF signal and demodulate data; it is a specification that is associated solely with the receiver<br />
system and is independent <strong>of</strong> the transmitter. As the signal propagates away from the transmitter,<br />
the power density <strong>of</strong> the signal decreases, this makes it more difficult <strong>for</strong> a receiver to detect the<br />
signal as the distance increases. Receivers with desirable sensitivity levels are able to increase<br />
the range over which surveillance or reception <strong>of</strong> meaningful signals is possible.<br />
The more common methods <strong>of</strong> specifying the sensitivity <strong>of</strong> a radio receiver include using SNR<br />
or Signal to Noise and Distortion ratio (SINAD). In the case <strong>of</strong> the SNR method, the sensitivity<br />
is realised by determining the input signal power necessary in order to achieve a specified SNR.<br />
The predetermined SNR is a function <strong>of</strong> the receiver application. For the experiments conducted<br />
here the SNR is set to be 10dB.<br />
It is worthwhile to note that, in addition to the basic per<strong>for</strong>mance <strong>of</strong> the receiver, the receiver's<br />
bandwidth can affect the SNR specification. It is found that the noise power decreases as the<br />
system bandwidth decreases. This implies that systems with smaller bandwidths collect less<br />
noise power [23].<br />
The SINAD measurement takes into account the degradation <strong>of</strong> the signal by unwanted<br />
contributions including noise and distortion. More specifically, SINAD is defined as the ratio <strong>of</strong><br />
the total signal power level, that being the signal, the noise and the distortion to the unwanted<br />
signal power, that being the noise and distortion.<br />
SINAD = 10log SND<br />
ND <br />
(4.5)<br />
where:<br />
SND<br />
ND<br />
is the combined signal, noise and distortion power level<br />
is the noise and distortion power level<br />
Similarly the sensitivity is assessed by determining the input level required in order to achieve a<br />
given figure <strong>of</strong> SINAD.<br />
In Figure 4.9, it is shown how the sensitivity <strong>of</strong> the TVRx and USRP varies. As the level <strong>of</strong> gain<br />
is increased, it is shown that a minimum signal <strong>of</strong> -105dBm is needed in order to produce the<br />
requisite 10dB <strong>of</strong> SNR.<br />
66
Figure 4.9: Variation <strong>of</strong> receiver sensitivity<br />
4.10 Dynamic Range, Minimum Detectable Signal and<br />
Gain Compression<br />
The standard operation <strong>of</strong> a receiver system is typically in the region where the output power is<br />
linearly proportional to the input power [3]. The constant <strong>of</strong> proportionality in this region is<br />
known as the conversion gain or conversion loss. It is this region <strong>of</strong> proportionality that is<br />
referred to as the dynamic range <strong>of</strong> a receiver. If the input power goes below the minimum<br />
acceptable signal, the minimum detectable signal (MDS), the noise effects dominate the receiver<br />
and results in nonlinear behaviour. This is due to the effects <strong>of</strong> the various sources <strong>of</strong> noise into<br />
the receiver including the components themselves.<br />
If the input power goes above the maximum allowable power <strong>of</strong> this region the receiver will<br />
display nonlinear characteristics as well. This behaviour may be as a result <strong>of</strong> damage to the<br />
receiver components at high power levels or due to gain compression or the generation <strong>of</strong><br />
spurious frequency components, due to device nonlinearities. These effects are undesirable as<br />
they may lead to increased losses, signal distortion and depending on the application possible<br />
interference with additional radio channels or services.<br />
67
Thus the definition <strong>of</strong> dynamic range can be summarised as:<br />
Maximum allowable signal power (dB) - Minimum detectable signal power (dB)<br />
In the first instance, the maximum allowable signal has been shown to be as a result <strong>of</strong> gain<br />
compression or saturation. Physically this effect is due to the limitation on the instantaneous<br />
power output imposed by the power supply used to bias the device [23]. In order to quantify the<br />
range over which the device is considered to be operating in the linear region, the 1dB<br />
compression point is defined. This is, the power level <strong>for</strong> which the output power <strong>of</strong> the device<br />
has deviated by 1dB from the ideal characteristic, <strong>for</strong> receiver systems this is usually specified in<br />
terms <strong>of</strong> the input power.<br />
Although 1dB compression points are most commonly used, 3dB and 10dB compression points<br />
are also used in some system specifications. The 1dB compression point is an important measure<br />
which can be used in the derivation <strong>of</strong> a receiver system's gain, bandwidth, noise figure and<br />
dynamic range [3]. In this instance, dynamic range is defined as: DR = P in, 1dB<br />
− MDS .<br />
This is shown in Figure 4.10.<br />
Figure 4.10: MDS, 1 dB compression point and dynamic range [3]<br />
The maximum allowable signal has been shown furthermore to be as a result <strong>of</strong> the generation <strong>of</strong><br />
spurious frequency components. This leads to an alternative definition <strong>of</strong> dynamic range, the<br />
spurious free dynamic range (SFDR). When a single frequency or tone is considered, generally<br />
68
the output will consist <strong>of</strong> harmonics <strong>of</strong> the frequency in the <strong>for</strong>m n 0<br />
Usually these harmonics do not affect the desired signal frequency.<br />
, where n = 0,1,2,<br />
This, however, is not the case when the signal is comprised <strong>of</strong> two or more closely spaced<br />
frequencies 1 and 2 . In particular, the third order intermodulation products,<br />
2 1<br />
− 2 and 2 2<br />
− 1 , are <strong>of</strong> interest. These terms are located near to the input<br />
frequencies and generally they are <strong>of</strong> relatively large magnitude, making them difficult to filter<br />
from the desired output and resulting in distortion <strong>of</strong> the output signal.<br />
Thus the third order intercept point (IP3 or TOI) is defined, as a measure <strong>of</strong> the intermodulation<br />
suppression. A high intercept point is indicative <strong>of</strong> a high suppression <strong>of</strong> undesired<br />
intermodulation products and is the fictitious point where the desired signal power and the third<br />
order signal powers are equal. In general the IP3 point occurs at about a 12dB - 15dB higher<br />
power level than the 1dB compression point. The IP3 point is an important measure <strong>of</strong> a systems<br />
linearity and is additionally specified in terms <strong>of</strong> the input power to the receiver.<br />
In this instance, dynamic range is defined as: DR sf<br />
= 2 3 IP3 − MDS .<br />
This is shown in Figure 4.11.<br />
Figure 4.11: Spurious Free Dynamic Range [3]<br />
69
Figure 4.12: Input power against output power, showing receiver dynamic range<br />
Figure 4.12 shows the variation <strong>of</strong> the output power with the power input to the receiver; from<br />
this we see as gain is increased from 0dB the MDS is determined at increasingly lower power<br />
levels. This is indicative <strong>of</strong> the increasing receiver sensitivity. Furthermore, as the gain setting<br />
increases we see three distinctive bands as each <strong>of</strong> the amplifier stages <strong>of</strong> the TVRx begin to run<br />
into saturation, thus the 1dB compression point begins to travel lower. This effectively indicates<br />
the reduction <strong>of</strong> the receiver's dynamic range at these high gain levels. The best achievable<br />
dynamic range <strong>of</strong> the receiver is noted as 62dB. This is in the region, mentioned in section 4.4,<br />
where GNU<strong>Radio</strong> most accurately modelled the TVRx amplifier stages.<br />
Figure 4.13 relates the measured dynamic range <strong>of</strong> 62dB to the theoretical SNR <strong>of</strong> the 12 bit<br />
ADC, which is calculated to be approximately 72dB using Equation 3.1. In practice however, the<br />
dynamic range available from the ADC is reduced as a result <strong>of</strong> thermal noise, reference noise,<br />
clock jitter, quantisation noise, etc. Figure 4.13 shows a 6dB loss in the ADC dynamic range as a<br />
result <strong>of</strong> noise, thus providing an effective 66dB <strong>of</strong> dynamic range. The minimum detectable<br />
signal from the TVRx, using the gain setting determined from Figure 4.12, was found to be 3dB<br />
above the ADC noise floor. As a result a further 3dB is lost from the ADC dynamic range as<br />
weaker signals from the TVRx will be noisy and not useful. Furthermore, the 1dB compression<br />
power output from the TVRx is found to be 1dB within the range that the ADC can extend to.<br />
Although this 1dB is available to the TVRx, stronger signals into the TVRx will begin to drive<br />
the amplifiers into saturation. There<strong>for</strong>e, although 66dB <strong>of</strong> dynamic range is available at the<br />
ADC, the TVRx is able to provide 62dB <strong>of</strong> dynamic range.<br />
70
Figure 4.13: TVRx and ADC dynamic range comparison<br />
4.11 Summary<br />
A number <strong>of</strong> the TVRx and USRP receiver system parameters as well as the GNU<strong>Radio</strong><br />
interface to the controllable aspects are examined. The receiver's frontend provides a static<br />
bandwidth <strong>of</strong> 6MHz and a tunable range between 50MHz and 800MHz, with a tuning step size<br />
as low as 31.25kHz. This is sufficient <strong>for</strong> the manipulation <strong>of</strong> FM Broadcast signals as the radar<br />
wave<strong>for</strong>m and provides further functionality and versatility to incorporate additional areas <strong>of</strong> the<br />
frequency spectrum if desired, i.e. the television video and sound carriers.<br />
The multiplexer implemented in the FPGA and the ability to directly control its operation<br />
through GNU<strong>Radio</strong> further extends the receiver's flexibility in its digital manipulation <strong>of</strong> signals<br />
through it. The noise characterisation <strong>of</strong> the receiver reveals a NF <strong>of</strong> 10dB, a sensitivity <strong>of</strong><br />
-105dB and a dynamic range <strong>of</strong> 62dB.<br />
71
Furthermore, it is observed that when the GNU<strong>Radio</strong> application is started, when the USRP is<br />
setup as a source and includes the occasion when the application is stopped and restarted, a<br />
spurious spike occurs at the beginning <strong>of</strong> the data set. This phenomena is unexplained but noted<br />
by the USRP developers, The effects <strong>of</strong> which are mitigated by simply cutting out that part <strong>of</strong> the<br />
data set.<br />
The following chapter will examine the suitability <strong>of</strong> FM broadcast signals as a candidate <strong>for</strong> the<br />
wave<strong>for</strong>m or part <strong>of</strong> a group <strong>of</strong> wave<strong>for</strong>ms used in the PCL radar. This is achieved by<br />
examination <strong>of</strong> the ambiguity plot and the variation <strong>of</strong> the instantaneous signal bandwidth.<br />
72
Chapter 5<br />
Ambiguity Plot Analysis<br />
5.1 Introduction<br />
In considering the use <strong>of</strong> broadcast FM signals <strong>for</strong> our PCL system, it is necessary to fully<br />
understand the nature <strong>of</strong> the signal parameters and the resulting effect and limitation on the<br />
system's per<strong>for</strong>mance. Range and doppler resolution in addition to range and doppler ambiguity<br />
are such parameters and they govern the ability to distinguish between two or more targets by<br />
virtue <strong>of</strong> their spatial or frequency differences and the ambiguity function has long been used to<br />
evaluate them [29] [24] [22].<br />
The nature <strong>of</strong> the transmitted wave<strong>for</strong>m determines these properties, <strong>for</strong> example, FM<br />
broadcasting uses the frequency range 88 – 108MHz [21] and the specified bandwidth is<br />
approximately 150kHz which gives a range resolution <strong>of</strong> 1–2km. This is further exemplified in<br />
the consideration <strong>of</strong> the nature <strong>of</strong> analogue television whereby the 64µs line repetition rate<br />
generates strong ambiguities [12] and consequently imposes a severe restriction and<br />
complication on its per<strong>for</strong>mance and applications.<br />
As can be imagined, the programme content and thus the instantaneous modulation <strong>of</strong> a<br />
particular broadcast FM signal varies with time. There<strong>for</strong>e, the ambiguity behaviour will vary as<br />
a function <strong>of</strong> the time and it is necessary to determine this behaviour.<br />
5.2 <strong>System</strong> Overview<br />
The receiving system used to capture and digitise the FM broadcast signals is described here.<br />
The system was situated on level 6 <strong>of</strong> the Menzies building at the University <strong>of</strong> Cape Town and<br />
detects and digitises broadcast signals originating from the Constantiaburg transmitter. It uses an<br />
antenna feeding the USRP, with the TVRx daughterboard plugged in and is thus in line with the<br />
project's low budget objectives. The daughterboard provides the down-conversion <strong>of</strong> the FM<br />
signal to the IF <strong>of</strong> 43.75MHz and is subsequently digitised on the USRP. The digitised signal is<br />
then further down converted to baseband and then decimated, using the standard USRP FPGA<br />
73
implementation. The data stream is then transported via USB and stored in the memory <strong>of</strong> a<br />
standard PC.<br />
Figure 5.1: Experiment setup to capture signals <strong>for</strong> ambiguity function analysis<br />
This choice <strong>of</strong> setup allows <strong>for</strong> the system to be flexible, as the TVRx module can be tuned to<br />
operate over the frequency range <strong>of</strong> interest and the captured data can be processed <strong>of</strong>f-line, as<br />
desired.<br />
We then compute the ambiguity function and the results <strong>of</strong> this computation are presented in a<br />
later section. Figure 5.1 above shows the experimental equipment and Figure 5.2 below<br />
represents this schematically.<br />
Figure 5.2: Block Diagram <strong>of</strong> experimental setup<br />
74
5.3 Algorithm Concept<br />
The purpose <strong>of</strong> this investigation is to determine the best achievable resolutions <strong>of</strong>fered by the<br />
FM broadcast signals and the variation <strong>of</strong> these signal properties with respect to time. It is<br />
important to understand this variation due to its effect on the overall range and doppler<br />
resolutions. For this process, knowledge <strong>of</strong> the relative positions <strong>of</strong> the transmitter, receiver and<br />
target are not required and thus the monostatic geometry is assumed in the following setup. This<br />
work is based upon the literature presented by Baker et al and by Howland et al and duplicates<br />
their research.<br />
As previously discussed, the nature <strong>of</strong> the transmitted wave<strong>for</strong>m determines these properties and<br />
is evaluated by computation <strong>of</strong> the ambiguity function, given by Equation 2.21.<br />
The output <strong>of</strong> a matched filter is represented by the ambiguity function and thus, computation <strong>of</strong><br />
the ambiguity function results in a three dimensional plot <strong>for</strong> which one axis is time delay, the<br />
second axis is Doppler frequency and the third is the normalised output power, computed by<br />
matched filtering the directly received transmitter signal. This processing is exemplified in the<br />
figure below.<br />
Figure 5.3: Signal processing algorithm[16]<br />
In the context <strong>of</strong> a PCL system, this signal is known to be the reference signal and is the<br />
wave<strong>for</strong>m that is correlated with the indirect target scattering in order to produce the output that<br />
is applied to the CFAR detectors, to obtain range and doppler in<strong>for</strong>mation <strong>of</strong> each target.<br />
This processing is written in discrete time notation as:<br />
∣ R R , f d ∣ 2 =∣<br />
N ∑ −1<br />
n=0<br />
2<br />
sns ∗ n−R R<br />
exp[ j 2 f d<br />
n/ N ]∣<br />
(4.6)<br />
75
A 0.1 second data sample <strong>of</strong> each FM Broadcast frequency <strong>of</strong> interest is digitised, with the<br />
USRP decimation rate set to the maximum value (256) this reduces the processing load and<br />
lowers the algorithm computation times. The effective sample rate is thus set to be 250kHz,<br />
which is well within the Nyquist criterion. The algorithm operating on each sample <strong>of</strong> data<br />
generates doppler shifted copies <strong>of</strong> the reference signal, each matched to a different target<br />
velocity.<br />
The implementation <strong>of</strong> the processing begins by rotating and then conjugating the samples <strong>of</strong> the<br />
reference signal sn , producing the values representing the time delay, s ∗ n−R R<br />
.<br />
These values are multiplied with the original reference signal and produce the result<br />
sn s ∗ n−R R<br />
. The Fourier trans<strong>for</strong>m <strong>of</strong> this result is determined and repeated <strong>for</strong> each<br />
range <strong>of</strong> interest.<br />
The USRP and GNU<strong>Radio</strong> were setup to digitise twenty successive 0.1 second samples, thereby<br />
providing 2 seconds <strong>of</strong> contiguous data <strong>for</strong> the analysis <strong>of</strong> the variation <strong>of</strong> range resolution with<br />
respect to time, <strong>for</strong> the transmission types <strong>of</strong> interest. The findings <strong>of</strong> this analysis follow.<br />
5.4 Results<br />
The ambiguity plots derived from the system above will be analysed in this section, illustrating<br />
the variability <strong>of</strong> the responses. Only the first set <strong>of</strong> plots and the associated table illustrating the<br />
variation <strong>of</strong> the signal bandwidth will be shown in this section.<br />
The ambiguity plots and the associated zero doppler and delay cuts as well as the tables <strong>of</strong> the<br />
remaining signals will be presented in appendix A, however a discussion <strong>of</strong> the results will be<br />
included in this section.<br />
The ambiguity plots shown are those <strong>of</strong> the six FM radio stations available from the<br />
Constantiaburg transmitter, namely these are (a) RAD5, the content on this channel is comprised<br />
mainly <strong>of</strong> rock, pop, RnB,hip hop and dance. This is dependant on the time <strong>of</strong> the day and week.<br />
Every hour, however, a five minute news bulletin is broadcast. (b) <strong>Radio</strong> 2000, the majority <strong>of</strong><br />
this content is a mixture <strong>of</strong> rock, jazz and talk. (c) Umhlobo, depending on the time <strong>of</strong> day<br />
broadcasts talk shows in the Xhosa medium, RnB and reggae. (d) Classic FM, which broadcasts<br />
an array <strong>of</strong> contemporary and classic works. (e) RGHP, <strong>of</strong>fers a selection <strong>of</strong> rock and country<br />
music. (f) RSGR, the dominant medium <strong>of</strong> the content that is broadcast is Afrikaans. Talk<br />
shows, pop, rock and country are all broadcast. (g) SAFM, which broadcasts news and talk<br />
programmes <strong>of</strong> various subjects and topic.<br />
Figure 5.4 below shows the ambiguity function <strong>for</strong> a RAD5 transmission. The signal content was<br />
comprised <strong>of</strong> rock. The ambiguity function plot shows a narrow defined peak and demonstrates<br />
fast fluctuating detail in the sidelobe regions. This indicates a wide bandwidth and due to the<br />
randomness in the signal modulation, exhibits noise-like characteristics and behaviour. which is<br />
consistent with what might be expected <strong>for</strong> rock music as compared to speech, which will be<br />
shown later. The zero range and doppler sections, in Figure 5.4a and Figure 5.4b below, show<br />
sidelobes to be usefully low with good per<strong>for</strong>mance in the range domain, which is reflective <strong>of</strong><br />
the high rate modulation in the rock wave<strong>for</strong>m. Sidelobe levels are good with about 27dB in the<br />
frequency domain and around 33dB <strong>for</strong> range.<br />
76
Figure 5.4: Ambiguity plot <strong>of</strong> Rad5 transmission<br />
Figure 5.4a: Zero doppler cut through ambiguity plot<br />
77
Figure 5.4b: Zero delay cut through ambiguity plot<br />
Upon analysis <strong>of</strong> the time dependant per<strong>for</strong>mance <strong>of</strong> this transmission, over a 2 second period,<br />
the behaviour is shown to be relatively consistent. This is illustrated in Table 5.1 below, which<br />
shows the bandwidth in kHz <strong>for</strong> twenty wave<strong>for</strong>m samples.<br />
Table 5.1: Bandwidth Variation <strong>of</strong> RAD5 wave<strong>for</strong>m<br />
Elapsed Time (s)<br />
Bandwidth (kHz)<br />
0.1 30<br />
0.2 50<br />
0.3 42<br />
0.4 55<br />
0.5 45<br />
0.6 52<br />
0.7 40<br />
0.8 43<br />
0.9 10<br />
1.0 32<br />
1.1 50<br />
1.2 47<br />
1.3 40<br />
78
1.4 40<br />
1.5 44<br />
1.6 50<br />
1.7 47<br />
1.8 43<br />
1.9 40<br />
2.0 42<br />
Average Bandwidth (kHz) 44.38<br />
The bandwidth is seen to vary in between 30kHz and 55 kHz, however, an anomaly <strong>of</strong> 10kHz is<br />
observed. This can be attributed to a break or pause in the modulating signal. The signal<br />
bandwidth averages 44.38kHz, making use <strong>of</strong> approximately 30% <strong>of</strong> the available bandwidth.<br />
Consequently, as the bandwidth is shown to be a function <strong>of</strong> time, the per<strong>for</strong>mance <strong>of</strong> the radar<br />
system will also be a function <strong>of</strong> time.<br />
Figure A.1 shows the ambiguity function <strong>for</strong> a <strong>Radio</strong> 2000 transmission.<br />
The ambiguity function plot again shows a defined peak, indicating a wave<strong>for</strong>m with noise like<br />
properties. In this instance, distinct peaks in the sidelobe regions are observed, and thus<br />
potential ambiguity. However, the zero range and doppler sections provide further insight and<br />
show the sidelobe levels are good with about 30dB in the frequency domain and around 30dB <strong>for</strong><br />
range and useful resolution in range and thus eliminating concerns over ambiguity.<br />
Table A.1 examines the time dependent bandwidth <strong>of</strong> the <strong>Radio</strong> 2000 transmission. Generally<br />
the values <strong>of</strong> bandwidth are a factor <strong>of</strong> two lower than that <strong>of</strong> the RAD5 transmission, this can be<br />
attributed to the nature <strong>of</strong> the modulating content. The variation swings between 0.5kHz to<br />
50kHz, which is much greater than that shown above and much more significant in its influence<br />
<strong>of</strong> the radar's per<strong>for</strong>mance. Additionally, the modulation bandwidth measured is 13% <strong>of</strong> that<br />
available.<br />
Figure A.2 , derived from a Classic FM transmission, shows a well defined ambiguity function<br />
peak, however,wider than those we have seen thus far, this can be attributed to the narrower<br />
bandwidth associated with this wave<strong>for</strong>m modulation.<br />
The zero range and doppler sections shown in Figures A.2a and A.2b demonstrate the range and<br />
doppler resolutions more clearly. The range resolution shows a slight degradation in<br />
per<strong>for</strong>mance, as is expected, due to the narrower bandwidth <strong>of</strong>fered. Sidelobes are seen close to<br />
the main peak but with reasonable levels at about 26dB in the range domain and around 40dB in<br />
the frequency domain. The peaks seen in the frequency domain away from the peak is measured<br />
at 33dB and thus does not impede the doppler per<strong>for</strong>mance.<br />
Table A.2 exhibits a variation in bandwidth from 0.6kHz to 32.5 kHz <strong>of</strong> the Classic FM<br />
transmission. The signal bandwidth averages 18.81kHz, making use <strong>of</strong> approximately 13% <strong>of</strong><br />
the available bandwidth.<br />
Figure A.3 shows the ambiguity function <strong>for</strong> a Umhlobo FM transmission.<br />
The ambiguity function plot again shows a narrow defined peak, <strong>of</strong>fering attractive range<br />
resolution and some detail in the sidelobe regions. In this analysis the signal content was<br />
comprised <strong>of</strong> kwaito music. This is loosely described as a South African variant <strong>of</strong> rock.<br />
Examining the zero range and doppler sections exhibit fast fluctuating detail associated with<br />
signals modulated with pop music. Indicating a more noise like wave<strong>for</strong>m structure and a wider<br />
bandwidth than that associated with rock, jazz and classic music. Doppler sidelobe levels are<br />
79
about 39dB and range sidelobe levels are additionally attractive with around 28dB, the range<br />
resolution <strong>of</strong>fered by the signal is useful.<br />
Table A.3 examines the time dependent bandwidth <strong>of</strong> the Umhlobo FM transmission. Generally<br />
the values <strong>of</strong> the bandwidth <strong>of</strong>fered are greater and the variation measured is less. The<br />
modulation bandwidth measured is on average 30% <strong>of</strong> that available.<br />
Figure A.4 shows the ambiguity function <strong>for</strong> a RGHP transmission.<br />
The ambiguity function plot again shows a defined peak, although narrower, relative to the other<br />
plots examined. This is indicative <strong>of</strong> the greater bandwidth <strong>of</strong>fered by the modulating music<br />
wave<strong>for</strong>m, in this instance RnB music. In the regions away from the main peak, we see a very<br />
clean structure with attractive sidelobe behaviour. This behaviour is more reflective <strong>of</strong> a random<br />
noise-like behaviour, which approaches the idealised “thumbtack” surface. An examination <strong>of</strong><br />
the zero range and doppler sections displays fast fluctuating detail and is associated with signals<br />
modulating music. Indicating a more random and noise like wave<strong>for</strong>m structure and a wider<br />
bandwidth than those examined thus far. Doppler sidelobe levels are about 32dB, which shows<br />
some measure <strong>of</strong> degradation relative to the previous signals, although not significant <strong>for</strong> our<br />
application. Range sidelobe levels however are good with around 28dB, with a very good range<br />
resolution <strong>of</strong>fered.<br />
Table A.4 examines the RGHP transmission. Generally the values <strong>of</strong> bandwidth are much<br />
improved over the alternative modulating types, averaging 63.6kHz, which translates to 42% <strong>of</strong><br />
that available.<br />
The signal from RSGR, in Figure A.5, was comprised <strong>of</strong> Afrikaans music. The peak <strong>of</strong> the<br />
ambiguity function is well defined but wider than those we have seen thus far <strong>for</strong> the alternative<br />
modulation types. The detail in the regions away from the main peak does not fluctuate as fast as<br />
the alternative signals examined thus far, this too is a function <strong>of</strong> the modulation present in this<br />
component <strong>of</strong> the wave<strong>for</strong>m. This behaviour is not reflective <strong>of</strong> pure noise-like behaviour,<br />
however, is consistent with the matched filter response that might be expected with a narrower<br />
bandwidth. The zero range and doppler sections shown below in Figures A.5a and A.5b<br />
demonstrate the range and doppler resolutions more clearly. The range resolution shows a<br />
degradation in per<strong>for</strong>mance, consistent with the decrease in the bandwidth associated with the<br />
signal. Sidelobe levels however are good with about 36dB in the frequency domain and around<br />
30dB <strong>for</strong> range.<br />
Table A.5 below demonstrates the behaviour <strong>of</strong> the RSGR transmission. Generally the values <strong>of</strong><br />
bandwidth are less than that <strong>of</strong> the rock, pop, and classic music modulating signals. The<br />
bandwidth is on average 11% <strong>of</strong> that available.<br />
The signal from SAFM in Figure A.6 was comprised <strong>of</strong> speech in a news broadcast. The peak <strong>of</strong><br />
the ambiguity function is extremely wide. This too is a function <strong>of</strong> the modulation present in this<br />
component <strong>of</strong> the wave<strong>for</strong>m and is consistent with the correlation that might be expected <strong>of</strong> a<br />
pause or break in speech. The zero range and doppler sections shown in Figures A.6a and A.6b<br />
demonstrate the range and doppler resolutions and sidelobe levels with about 40dB in the<br />
frequency domain and per<strong>for</strong>ming extremely poorly in range. This result is indicative <strong>of</strong><br />
ambiguous per<strong>for</strong>mance as the bandwidth <strong>of</strong>fered is very low, thus the potential resolution <strong>of</strong> the<br />
radar would be not at all useful in this instant.<br />
In Table A.6, the bandwidth is seen to vary from 1.4kHz to 47.5kHz. This average bandwidth<br />
used is 13% <strong>of</strong> that available. This is consistent with the analysis expected <strong>of</strong> a speech (news<br />
broadcast) transmission.<br />
80
All the bandwidths presented here are calculated from the -3dB width <strong>of</strong> the zero doppler<br />
section, through the ambiguity plot. It is evident that there is a great deal <strong>of</strong> variation in the<br />
per<strong>for</strong>mance <strong>of</strong> the available wave<strong>for</strong>ms to be exploited. This variation is dependent on the type<br />
<strong>of</strong> transmission, its content and the time <strong>of</strong> transmission. This can be understood, <strong>for</strong> example, if<br />
there is a long pause on a channel or with speech, the signal spectral content is reduced, thus<br />
range resolution in<strong>for</strong>mation will be degraded or lost.<br />
This is illustrated in Figure 5.5a and Figure 5.5b, which compares the range resolution against<br />
time <strong>for</strong> the different transmissions.<br />
Figure 5.5a: Variation <strong>of</strong> signal range resolution with time<br />
81
Figure 5.5b: Variation <strong>of</strong> signal range resolution with time<br />
Here we see that RSGR and SAFM (talk show) show a high degree <strong>of</strong> temporal variability in<br />
range resolution compared to the music channels as is expected, with the resolution degrading in<br />
the order <strong>of</strong> hundreds <strong>of</strong> kilometres. The classical music does not show as much variation, thus<br />
demonstrating an increased per<strong>for</strong>mance. However, the pop and dance channels exhibit the least<br />
variation, with resolution varying between 1km and 15km.<br />
Thus the signal per<strong>for</strong>mance has quite a considerable variation and the effect on overall system<br />
per<strong>for</strong>mance will require careful consideration.<br />
5.5 Summary<br />
From the results it is evident that the measured FM Broadcast signals due to the randomness in<br />
their modulation, exhibit noise like characteristics and behaviour. This is observed as the<br />
ambiguity function plots can be approximated by the ideal thumbtack ambiguity function,<br />
there<strong>for</strong>e providing the radar with excellent range and doppler in<strong>for</strong>mation. However, it is<br />
additionally clear that the ambiguity function depends largely on the modulating <strong>for</strong>mat.<br />
82
Thus there is a variation in per<strong>for</strong>mance between the pop music and dance music channels and<br />
the rock and classical music channels which exhibit a slightly degraded per<strong>for</strong>mance. Speech<br />
content shows the poorest per<strong>for</strong>mance though, as the per<strong>for</strong>mance does vary significantly with<br />
time, especially during pauses in words due to the low spectral content. This variation in<br />
per<strong>for</strong>mance does have a significant impact on the radar's potential detection ability and will be<br />
one that does vary with time, as seen from Equation 2.6 and Table A.6, this variation translates<br />
to a change in processing gain <strong>of</strong> 15dB.<br />
Table 5.2 below summarises the ambiguity function per<strong>for</strong>mance <strong>of</strong> the measured signals.<br />
Signal<br />
Table 5.2: Summary <strong>of</strong> wave<strong>for</strong>m ambiguity function per<strong>for</strong>mances<br />
Range Resolution<br />
(km)<br />
Effective<br />
Bandwidth<br />
(kHz)<br />
Peak range<br />
Sidelobe level<br />
(dB)<br />
Peak doppler<br />
Sidelobe level<br />
(dB)<br />
RAD5 3.38 44.3 - 30 - 27<br />
<strong>Radio</strong> 2000 7.94 18.9 - 18 - 34<br />
Classic FM 7.98 18.8 - 30 - 33<br />
Umhlobe 3.33 45.0 - 31 - 25<br />
RGHP 2.36 63.6 - 30 - 37<br />
RSGR 9.32 16.1 - 18 - 36<br />
SAFM 7.54 19.9 - 20 - 40<br />
Average 5.98 32.37 -25.29 -33.14<br />
83
Chapter 6<br />
Oscillator Stability<br />
6.1 Introduction<br />
A significant limiting factor in the per<strong>for</strong>mance <strong>of</strong> radar applications is the ability <strong>of</strong> the radar's<br />
frequency reference to maintain timing accuracy.<br />
Reviewing the theory <strong>of</strong> previous chapters, it is understood that in radar, the range <strong>of</strong> the target is<br />
related to time.<br />
R = c⋅ d<br />
2<br />
Where:<br />
R<br />
d<br />
is the range <strong>of</strong> the target, from the radar receiver.<br />
is the round trip time delay <strong>of</strong> the signal, between the transmitter and receiver.<br />
Thus, a radar’s range accuracy is directly proportional to the error in the timing signal [24] , as<br />
an error in d is directly proportional to an error in R .<br />
In a doppler radar, the signal received by the radar from a moving target differs in frequency<br />
from the transmitted frequency and thus differs in phase too, by an amount that is proportional to<br />
the radial component <strong>of</strong> the velocity relative to the radar.<br />
This shift is is given by:<br />
f d<br />
= 2v r<br />
<br />
As a result, the velocity <strong>of</strong> the target and the radar frequency are primary determinants <strong>of</strong> the<br />
phase noise requirements [27]. A fast moving target will cause a large doppler frequency, and<br />
thus will require a low phase noise at frequencies far from the carrier. In quartz crystals most <strong>of</strong><br />
the noise power lies close to the carrier [3] [23], consequently, low phase noise close to the<br />
carrier, required <strong>for</strong> a slow moving target, is not easily achieved [24].<br />
84
Moreover, coherence, in radar applications is defined as whether the phase relationship (at the<br />
carrier frequency) between successive pulses are known and stable [24]. If a radar is coherent,<br />
then M successive pulses, from a non-fluctuating target, can be added. This summing will<br />
improve the radar's SNR by a factor <strong>of</strong> M. However, if the receiver is not coherent, these pulses<br />
will not add in phase and summing will not improve the SNR. Refer to chapter 2.6.4 <strong>for</strong> a more<br />
thorough treatment <strong>of</strong> this point.<br />
Furthermore, as the radar's probability <strong>of</strong> detection is directly related to the SNR, degradation <strong>of</strong><br />
the radar's coherence will decrease the SNR. As a result, in coherent radar and thus PCL, the<br />
target detection capability decreases with an increase in phase noise <strong>of</strong> the frequency reference.<br />
In a bistatic or multistatic radar configuration, the complexity <strong>of</strong> synchronisation and coherence<br />
is further increased. The local oscillators in all the receivers, need to be synchronised not only in<br />
frequency but the phase <strong>of</strong>fsets between them needs to be known as well [27]. The degree to<br />
which these phase <strong>of</strong>fsets stay constant during target detection will influence the level <strong>of</strong><br />
coherence. In a passive bistatic radar system the echo signal is correlated with a bank <strong>of</strong> doppler<br />
shifted replicas <strong>of</strong> the transmitted signal, over a predetermined sample period, in order to achieve<br />
the necessary signal processing gain and <strong>for</strong> target detection. Thus the phase coherence is only<br />
required during this sample period. The degree <strong>of</strong> coherence can be thought <strong>of</strong> as how stable the<br />
local oscillator clock, in the TVRx, remains.<br />
Assuming a target velocity <strong>of</strong> 150m/s, a broadcast FM transmission frequency <strong>of</strong> 100MHz and<br />
using Equation 2.17, a target will produce a doppler shift <strong>of</strong> 100Hz. In order to qualify the results<br />
<strong>of</strong> this chapter a 5% error in the calculated doppler shift is set as the maximum acceptable limit<br />
<strong>of</strong> frequency shift. Thus a frequency drift <strong>of</strong> no greater than 5Hz is set as the requirement <strong>of</strong> the<br />
system.<br />
6.2 Quartz Crystal Oscillators<br />
The local oscillator in the TVRx daughterboard <strong>of</strong> the USRP is derived from a 4MHz quartz<br />
crystal oscillator. For this reason discussion will focus on the properties <strong>of</strong> crystal oscillators and<br />
the associated per<strong>for</strong>mance parameters <strong>of</strong> interest. These parameters <strong>of</strong> interest include accuracy,<br />
reproducibility and stability, although, oscillator stability will dominate the discussion.<br />
Accuracy, in the case <strong>of</strong> frequency, is a measure <strong>of</strong> how well the frequency source relates to the<br />
definition <strong>of</strong> one second. Reproducibility, is a measure <strong>of</strong> how well a number <strong>of</strong> sources agree in<br />
frequency as they are adjusted, this is more relevant as a factory acceptance test. Stability is a<br />
measure <strong>of</strong> how well a frequency source is able to generate a given frequency over some<br />
measure <strong>of</strong> time, once it has been set. Thus the difference between the frequency at one moment<br />
in time and another moment is called stability [19] and is usually given <strong>for</strong> a number <strong>of</strong> time<br />
periods, ranging between seconds and years.<br />
Crystal oscillators are widely used as frequency sources and find application in a wide variety <strong>of</strong><br />
areas from wristwatches to complex and elaborate instruments found in laboratories. These<br />
oscillators provide good per<strong>for</strong>mance at a reasonable price and dominate the field <strong>of</strong> frequency<br />
sources [19].<br />
The quartz crystal in the oscillator resonates mechanically and the associated oscillations <strong>of</strong> the<br />
resonator have to be sensed. This is done by taking advantage <strong>of</strong> the piezoelectric effect. The<br />
piezoelectric effect is observed when a physical compression <strong>of</strong> the crystal generates a voltage<br />
across the crystal. Conversely, the application <strong>of</strong> an external voltage across the crystal causes the<br />
crystal to either expand or contract depending on the polarity <strong>of</strong> the voltage.<br />
85
Limitations in stability stem mainly from oscillator drift and ageing effects. Further discussion<br />
regarding this will follow in section 6.2.1.<br />
6.2.1 Ageing and Temperature <strong>of</strong> Crystals<br />
The most influential factors affecting oscillator per<strong>for</strong>mance are ageing and temperature. There<br />
is a temperature dependence <strong>of</strong> the quartz crystal that affects its resonance frequency, and there<br />
is also instability <strong>of</strong> the resonance frequency due to ageing [19]. Drift is due to ageing plus<br />
changes in the environment and other factors external to the oscillator, thus oscillator drift is<br />
defined as the systematic change in frequency with time <strong>of</strong> an oscillator. Ageing, on the other<br />
hand, is defined as the systematic change in frequency due to physical changes in the oscillator<br />
[27]. This ageing rate decreases with time.<br />
Ageing is a characteristic common to crystal oscillators and can be approximated as a linear<br />
change in resonance frequency with time. In general, the drift is negative, meaning that the<br />
resonance frequency decreases. This decrease is indicative <strong>of</strong> an increase in the size <strong>of</strong> the<br />
crystal. Possible causes <strong>of</strong> this phenomenon are contamination on the surface <strong>of</strong> the crystal,<br />
variations in the electrodes or the metallic plating, the movement <strong>of</strong> loose surface material from<br />
grinding and etching or changes in the internal crystal structure.<br />
Furthermore, the continuous expansion and contraction <strong>of</strong> the crystal, due to the piezoelectric<br />
effect, may be the cause <strong>of</strong> or at least exacerbating these effects. Improvements in crystal holder<br />
design combined with clean vacuum enclosures, have led to a reduction in oscillator ageing [19].<br />
The temperature effects on crystal per<strong>for</strong>mance are as a result <strong>of</strong> slight changes in the elastic<br />
properties <strong>of</strong> the crystal. Special crystal cutting techniques and variations known as<br />
crystallographic orientations minimize this effect <strong>of</strong> temperature over a range <strong>of</strong> temperatures.<br />
Temperature coefficients <strong>of</strong> less than one part in 100 million per degree <strong>of</strong> temperature are<br />
possible.<br />
Crystal oscillators must be carefully designed if very high frequency stabilities are desired. If<br />
large environmental temperature fluctuations are to be tolerated, the crystals themselves are<br />
enclosed in electronically regulated ovens which maintain a constant temperature. These<br />
oscillators are known as Oven Controlled Crystal Oscillators (OCXO).<br />
6.3 Characterisation <strong>of</strong> stability<br />
The techniques associated with the specification <strong>of</strong> stability are presented in the following<br />
sections and in order to appreciate this, a presentation <strong>of</strong> the relationship between frequency and<br />
phase will, additionally, be made.<br />
It is assumed that the average output frequency <strong>of</strong> a precision oscillator is determined by a<br />
narrow band circuit (crystal oscillator), so that the signal can be approximated as a sine wave<br />
[19] and thus the signal output is generalised by the following <strong>for</strong>mula.<br />
V t = [V 0<br />
t ]sin [2 v 0<br />
t t ]<br />
(6.1)<br />
86
where:<br />
V 0<br />
t<br />
v 0<br />
t<br />
is the nominal peak voltage amplitude<br />
is the deviation <strong>of</strong> the amplitude from the nominal<br />
is the nominal fundamental frequency<br />
is the deviation <strong>of</strong> the phase from the nominal<br />
This can be furthermore generalised and rewritten as the t in precision oscillators is<br />
generally ignored.<br />
V t = V 0<br />
sin [2v 0<br />
t t]<br />
= V 0<br />
sin t <br />
(6.2)<br />
where:<br />
t<br />
is the total oscillator phase<br />
Here we note that due to a change in amplitude, v , over an associated change in time, t ,<br />
we able to determine the average frequency over this period. Furthermore, in the limit as t<br />
approaches zero an instantaneous frequency can be measured. This is however, in practice, not<br />
possible as it requires an infinite bandwidth. As a result we are only able to measure the<br />
frequency that has been averaged over a time period t .<br />
Furthermore, the frequency <strong>of</strong> a signal is additionally related to the rate <strong>of</strong> change <strong>of</strong> its phase.<br />
The instantaneous angular frequency is defined as the time derivative <strong>of</strong> the total oscillator<br />
phase. Thus,<br />
t = d dt [2v 0 t t]<br />
(6.3)<br />
= 2 v 0<br />
d <br />
dt<br />
The instantaneous frequency is then written as:<br />
V t = v 0<br />
1<br />
2<br />
d <br />
dt<br />
(6.4)<br />
87
For precision oscillators, the second term on the right hand side is quite small and useful to<br />
define the fractional frequency.<br />
yt = vt − v 0<br />
v 0<br />
= 1 d <br />
2 v 0<br />
dt<br />
= dx<br />
dt<br />
(6.5)<br />
where:<br />
xt = t<br />
2 v 0<br />
(6.6)<br />
xt is an expression <strong>of</strong> the phase in units <strong>of</strong> time and is thus representative <strong>of</strong> the total<br />
cumulative time deviation <strong>of</strong> the reference clock due to instability.<br />
This quantity can be additionally determined by integrating the fractional deviation<br />
respect to time.<br />
yt , with<br />
However, the bandwidth limitations that apply to the measurement <strong>of</strong> instantaneous frequency<br />
are additionally applicable to instantaneous phase measurement. There<strong>for</strong>e, the average<br />
fractional frequency is defined as:<br />
yt =<br />
xt − xt <br />
<br />
(6.7)<br />
xt and xt are the respective time deviations measured. is known as the<br />
sampling or averaging time. This is illustrated in the Figure 6.1.<br />
Figure 6.1: Determining the fractional frequency<br />
88
In Figure 6.2, v(t) represents a stable reference oscillator and w(t) represents an oscillator with<br />
instabilities. If viewed over a timescale smaller than t 1 , w(t) is unstable relative to v(t).<br />
However, above this time the oscillators are in phase and as a result, the instability <strong>of</strong> w(t) would<br />
have gone undetected if time measurements greater than t 1 were made. Furthermore, over the<br />
period t 1 to t 2 , although no instabilities are apparent, when analysed over a smaller time<br />
period, instabilities are noted.<br />
Figure 6.2: Dependence <strong>of</strong> stability measurement on time window<br />
6.3.1 Measurement <strong>of</strong> Stability<br />
A variety <strong>of</strong> methods and their variations exist in order to measure the stability <strong>of</strong> frequency<br />
sources. The methods most prominent in the literature will be reviewed in this section.<br />
Frequency stability can be measured by taking a reasonably large number <strong>of</strong> successive readings<br />
<strong>of</strong> the frequency <strong>of</strong> the device to be evaluated. Each reading, measured in hertz is obtained by<br />
averaging the output frequency <strong>for</strong> some specified time period. Furthermore, variations in the<br />
readings <strong>of</strong> measured frequency might be expected to average out if observed <strong>for</strong> long enough.<br />
This however, is not always the case [19].<br />
The results are then expressed, <strong>of</strong>ten, as a relative or fractional value, given by the following<br />
<strong>for</strong>mula.<br />
=[<br />
F f actual − f ]<br />
nameplate<br />
(6.8)<br />
f nameplate<br />
The nameplate frequency refers to the expected oscillator operating frequency. If an oscillator<br />
operated exactly at its nameplate frequency, it would be considered a perfect frequency source.<br />
However, in practice there is some frequency error and thus a difference f between the<br />
actual and nameplate frequencies. The value <strong>of</strong> the fractional frequency will become negative if<br />
the measured value is below that <strong>of</strong> the nominal frequency.<br />
89
The process <strong>of</strong> dividing the f by the nameplate frequency, normalises the error and allows<br />
frequency stability to be directly manipulated as it is independent <strong>of</strong> the actual operating<br />
frequency <strong>of</strong> the oscillator being discussed. For example this makes it possible to compare the<br />
stability <strong>of</strong> a 10MHz oscillator to a 10kHz oscillator without any additional in<strong>for</strong>mation.<br />
As a further example, a 1MHz oscillator that is higher in its frequency by 1Hz will have a<br />
fractional frequency equal to 10 -6 or 1 Part Per Million (PPM). The same 1Hz error <strong>for</strong> a 10MHz<br />
oscillator is a smaller part <strong>of</strong> the nominal frequency and gives us a smaller relative frequency<br />
equal to 10 -7 or 0.1PPM. Thus, this division <strong>of</strong> f by the nominal (nameplate) frequency,<br />
removes the concern <strong>of</strong> weather the nameplate frequency is 1 MHz or 10MHz and is extended to<br />
any other frequency <strong>of</strong> interest.<br />
In addition to this, this notation allows the measurement <strong>of</strong> the oscillator output directly or the<br />
use <strong>of</strong> a divided version <strong>of</strong> the oscillator. For example, an oscillator with a 1MHz output can<br />
have its output divided to 1Hz without changing the result <strong>of</strong> its stability. This attribute provides<br />
the ability to deal with sources <strong>of</strong> any frequency by using dividers to make the measurement<br />
problems more manageable, as it is much easier to deal with lower frequencies, and the final<br />
results are the same.<br />
In describing the second measure <strong>of</strong> frequency stability, it is useful to mention the relationship<br />
between the time and frequency domains <strong>of</strong> a signal.<br />
This is understood by use <strong>of</strong> the continuous Fourier trans<strong>for</strong>m pair, defined as:<br />
and<br />
∞<br />
X f = ∫ xt exp[− j 2 f t]dt<br />
−∞<br />
x t = 1 ∫ ∞<br />
X f exp[ j 2 f t]df<br />
2 <br />
−∞<br />
(6.9)<br />
(6.10)<br />
The frequency domain representation <strong>of</strong> the signal xt , the total cumulative time deviation, is<br />
also known as its power spectrum and provides an indication <strong>of</strong> the distribution <strong>of</strong> the signal's<br />
power with frequency. The power density spectrum, S x<br />
f , can then be obtained by<br />
normalising the power spectrum, such that the total area under the curve is equated to unity [30].<br />
Furthermore, the power spectrum contains frequency components <strong>for</strong> both phase and amplitude<br />
fluctuations and when dealing with precision oscillators, the mean squared value <strong>of</strong> the<br />
amplitude fluctuations are small enough to be neglected. The spectral density, can be then further<br />
scaled to represent the phase power density spectrum, S <br />
f . Finally, S y<br />
f represents<br />
the spectral density <strong>of</strong> the instantaneous fractional frequency fluctuations yt .<br />
These spectral densities are related in the following manner:<br />
S y<br />
f = f 2<br />
v 0<br />
2<br />
S <br />
f <br />
(6.11)<br />
S x<br />
f =<br />
1<br />
2 v 0<br />
2<br />
S f <br />
S <br />
f = 2 f 2 S <br />
f <br />
90
The third measure <strong>of</strong> frequency stability is to use the sample variance, y 2 , <strong>of</strong> the<br />
fractional frequency fluctuations. This variance determines the extent <strong>of</strong> the variation <strong>of</strong> the<br />
oscillator's average frequency between two adjacent measurement intervals. If the time or<br />
frequency fluctuations between a pair <strong>of</strong> oscillators is measured, a process whereby N values <strong>of</strong><br />
the fractional frequency y i are measured over a time and measurements are repeated after<br />
an interval <strong>of</strong> time T . There is a dead time between each frequency measurement, if the<br />
measurement repetition interval is greater than the averaging time and is given by T − , this<br />
is illustrated in Figure 6.3.<br />
The N sample variance can be then computed,<br />
〈 2 1<br />
y<br />
N ,T , 〉 =<br />
〈 ∑ y<br />
N − 1 n<br />
− 1 ∑<br />
n = 1 N k = 1<br />
N<br />
N<br />
2<br />
y k<br />
<br />
〉<br />
(6.12)<br />
Figure 6.3: measurement process [30]<br />
The angle brackets in Equation 6.12 denote the infinite time average. Frequently, however, this<br />
result does not converge as N ∞ . This is as a result <strong>of</strong> noise processes in the oscillator that<br />
diverge at low Fourier frequencies. Thus the precision <strong>of</strong> the estimation <strong>of</strong> the variance, does not<br />
simply improve as the sample size is increased. For this reason, the Allan variance is preferred as<br />
a measure <strong>of</strong> stability and is shown to converge rapidly as the sample size is increased [30]. A<br />
low Allan variance is a characteristic <strong>of</strong> a clock with good stability <strong>of</strong> the measured period.<br />
The Allan variance is described as [30]:<br />
y 2 = 〈 1 2 yt − y t 2 〉<br />
(6.13)<br />
91
This equation can be stated equivalently in terms <strong>of</strong> phase data as [30]:<br />
where:<br />
2 y<br />
=<br />
〈 1 xt 2 − 2xt xt<br />
2<br />
2<br />
2〉<br />
(6.14)<br />
xt is the measured phase difference. This can still furthermore, be manipulated in order to<br />
operate on a discrete time data set as [30]:<br />
y 2 ≃<br />
1<br />
2 N − 2 2<br />
N − 2<br />
∑ x i 2<br />
− 2x i 1<br />
x i<br />
2<br />
i = 1<br />
(6.15)<br />
6.4 <strong>System</strong> Overview<br />
This section will describe the experimental test setup. The test signal is generated by a Hewlett<br />
Packard 8656B signal generator, with the output frequency set to 100MHz. This signal is split<br />
and sent to the USRP with two TVRx daughterboard modules plugged in and running<br />
simultaneously. This is representative <strong>of</strong> the setup that will be expected <strong>of</strong> the PCL receiver with<br />
independent TVRx modules <strong>for</strong> the reference channel and the channel monitoring target returns.<br />
The independent signals are then digitised by the USRP, synchronisation <strong>of</strong> the ADCs is ensured<br />
as they are clocked by the same master clock running the USRP. The digitised signals are then<br />
written to file on a PC <strong>for</strong> <strong>of</strong>fline processing <strong>of</strong> the data. The functional block diagram <strong>of</strong> the<br />
system is shown in Figure 6.4 and a photograph <strong>of</strong> the test equipment is shown in Figure 6.5.<br />
Figure 6.4: Block diagram <strong>of</strong> experimental setup<br />
92
Figure 6.5: Experiment setup <strong>for</strong> stability tests<br />
GNU<strong>Radio</strong> was configured such that the signals from tuner 1 and tuner 2 are written to separate<br />
files on the PC, allowing <strong>for</strong> greater flexibility in post processing. The experiment was<br />
conducted with three variations <strong>of</strong> the decimation rate, that being 256, 128, and 64. This<br />
translates to data rates varying between 250ksps and 1Msps.<br />
Furthermore, the system was setup in order to determine the oscillator stability over a 30 minute<br />
period. This was achieved by taking successive readings <strong>of</strong> the frequency output by each tuner<br />
over the 30 minute period. Each reading, was obtained by averaging the output frequency <strong>for</strong> a 1<br />
second time gate, every minute over the 30 minute test time.<br />
The resolution <strong>of</strong> the measured frequency is given by:<br />
f = 1 T<br />
(6.16)<br />
Where T is the time gate over which the frequency is averaged.<br />
Furthermore, the phase variation between the signals at the measured intervals was determined<br />
and used in order to verify and confirm the frequency variation that had been determined. A set<br />
<strong>of</strong> fractional frequencies describing the behaviour <strong>of</strong> the stability <strong>of</strong> the oscillator were then<br />
obtained. These results are used in order to generate the Allan deviation <strong>of</strong> the oscillators.<br />
93
6.5 Results<br />
The following section will present the results <strong>of</strong> the tests conducted, using the experimental<br />
setup outlined in the above section.<br />
In Figure 6.6, the fractional frequency is observed over the 30 minute time frame. As expected<br />
the crystal ageing effect is gradually reduced over the period, showing a variation from 2.3 to 0.8<br />
parts per 10 million. The large separation initially can be attributed to the innate crystal<br />
oscillator behaviour described as the warm up effect, which is caused by the rise in temperature<br />
<strong>of</strong> the oscillator from the time the oscillator and instrument is turned on until the time a stable<br />
operating temperature is reached. The temperature rise is brought about directly by power<br />
dissipation in the oscillator or indirectly by the generation <strong>of</strong> heat in the circuitry <strong>of</strong> the<br />
instrument. Additionally, the Allan variance <strong>of</strong> these fractional frequency deviations is shown in<br />
Figure 6.7.<br />
Figure 6.6: Fractional frequency variation <strong>of</strong> the daughterboards<br />
The Allan variance determines the extent <strong>of</strong> the variation <strong>of</strong> the oscillator's average frequency<br />
between two adjacent measurement intervals. Thus, although we observe an overall decline in<br />
the fractional frequency to 0.8 parts per 10 million, the magnitude <strong>of</strong> the variation between<br />
successive measurements is less severe. The oscillation <strong>of</strong> the variation indicates movement <strong>of</strong><br />
the oscillator frequencies towards and away from each other. The largest measure <strong>of</strong> the<br />
instantaneous drift is 4Hz. This oscillation can be attributed to physical attributes <strong>of</strong> the crystal,<br />
losses in the oscillator circuitry, vibration or temperature effects.<br />
94
6.6 Summary<br />
Figure 6.7: Allan Variance <strong>of</strong> the fractional frequencies<br />
In this chapter it is shown that a critical limiting factor in the per<strong>for</strong>mance <strong>of</strong> radar applications<br />
is the ability <strong>of</strong> the radar's frequency reference to maintain timing accuracy.<br />
The local oscillator in the TVRx daughterboard <strong>of</strong> the USRP is a 4MHz quartz crystal oscillator.<br />
It is <strong>for</strong> this reason that the properties <strong>of</strong> crystal oscillators and the associated per<strong>for</strong>mance<br />
parameters <strong>of</strong> interest were investigated. These parameters include accuracy, reproducibility and<br />
stability. It was found that crystal oscillators provide good per<strong>for</strong>mance at a reasonable price and<br />
dominate the field <strong>of</strong> frequency sources [19].<br />
Furthermore, the techniques associated with the specification <strong>of</strong> stability were presented,<br />
including measurement <strong>of</strong> the frequency deviation and the Allan variance. These tests concluded<br />
that the crystal ageing effect is reduced over the measurement period <strong>of</strong> thirty minutes, showing<br />
a fractional frequency variation from 2.3 to 0.8 parts per 10 million and a frequency drift no<br />
greater than 4Hz. This falls within the maximum allowance <strong>of</strong> 5Hz, calculated in section 6.1.<br />
95
Chapter 7<br />
Conclusions and Recommendations<br />
This dissertation provides a comprehensive discussion around bistatic radar with specific<br />
reference to PCL. Various uses <strong>of</strong> these radar systems, such as military and civilian applications<br />
have been presented, as well as the advantages and disadvantages <strong>of</strong> the system have been<br />
discussed in detail, and make this type <strong>of</strong> system desirable <strong>for</strong> our applications. However, at the<br />
time <strong>of</strong> writing this dissertation various aspects <strong>of</strong> the overall project were still under<br />
investigation and as such explicit system requirements were not available, thus, an exhibition<br />
highlighting existing literature and work as well as an examination <strong>of</strong> the various per<strong>for</strong>mance<br />
metrics is made.<br />
In particular the ambiguity function is examined and the per<strong>for</strong>mance <strong>of</strong> commercial FM radio<br />
broadcasts as the radar wave<strong>for</strong>m is determined and summarised by Table 5.2, reproduced here.<br />
Signal<br />
Table 5.2: Summary <strong>of</strong> wave<strong>for</strong>m ambiguity function<br />
Range Resolution<br />
(km)<br />
Effective<br />
Bandwidth<br />
(kHz)<br />
Peak range<br />
Sidelobe level<br />
(dB)<br />
Peak doppler<br />
Sidelobe level<br />
(dB)<br />
RAD5 3.38 44.3 - 30 - 27<br />
<strong>Radio</strong> 2000 7.94 18.9 - 18 - 34<br />
Classic FM 7.98 18.8 - 30 - 33<br />
Umhlobe 3.33 45.0 - 31 - 25<br />
RGHP 2.36 63.6 - 30 - 37<br />
RSGR 9.32 16.1 - 18 - 36<br />
SAFM 7.54 19.9 - 20 - 40<br />
Average 5.98 32.37 -25.29 -33.14<br />
The analysis <strong>of</strong> the FM broadcast signal ambiguity plots reveal the attractive per<strong>for</strong>mance <strong>of</strong> the<br />
signals as they in general tend toward the idealised thumbtack. However, this is greatly<br />
dependant upon the instantaneous modulating content, revealing highly degraded per<strong>for</strong>mance in<br />
the case <strong>of</strong> a signal with low spectral content such as speech with many pauses and breaks.<br />
Furthermore, the SDR paradigm and technology is examined, with discussion around the design<br />
considerations. The USRP, the TVRx daughterboard and GNU<strong>Radio</strong> are examined further as a<br />
potential receiver and development environment, in this light.<br />
96
GNU<strong>Radio</strong>, is found to be a powerful toolkit that implements numerous complex signal<br />
processing elements and additionally interfaces with the USRP that provides flexible and<br />
powerful computing capabilities. Thus combining to provide a powerful and widely versatile<br />
toolset <strong>for</strong> the development <strong>of</strong> a variety <strong>of</strong> radio devices.<br />
The system meets the low cost ambitions costing just over US$500.00 <strong>for</strong> the USRP<br />
motherboard and a single daughterboard. The receiver's frontend provides a static bandwidth <strong>of</strong><br />
6MHz and a tunable range between 50MHz and 800MHz. The noise characterisation <strong>of</strong> the<br />
receiver reveals a NF <strong>of</strong> 10dB, a sensitivity <strong>of</strong> -105dB and a dynamic range <strong>of</strong> 62dB.<br />
Finally, the investigation into the stability <strong>of</strong> the daughterboard frontend oscillators due to ageing<br />
effects is shown to be steady through examination <strong>of</strong> the fractional frequency variation as well as<br />
the Allan variance, showing a fractional frequency variation from 2.3 to 0.8 parts per 10 million<br />
and a frequency drift no greater than 4Hz. This falls within the maximum allowance <strong>of</strong> 5Hz,<br />
calculated in section 6.1.<br />
In order to further increase the per<strong>for</strong>mance <strong>of</strong> the system, investigation into the use <strong>of</strong> multiple<br />
wave<strong>for</strong>ms is recommended. The bandwidth <strong>of</strong> the TVRx lends itself to this. Furthermore, the<br />
effect <strong>of</strong> multiple USRP receivers and the effect <strong>of</strong> location and thus multiple simultaneous<br />
locations can be investigated providing a richer in<strong>for</strong>mation source <strong>for</strong> the relevant signal<br />
processing.<br />
Special attention should be paid to timing and synchronisation effects and the examination <strong>of</strong> the<br />
“cross ambiguity function” modelling the bistatic geometry. In addition to this, the versatility <strong>of</strong><br />
GNU<strong>Radio</strong> can be further explored, using the inbuilt signal processing blocks to improve the end<br />
to end per<strong>for</strong>mance <strong>of</strong> the PCL system. i.e. The support and per<strong>for</strong>mance <strong>of</strong> adaptive filtering<br />
and various pre-correlation signal conditioning techniques may be explored.<br />
97
Bibliography<br />
[1] Asad A. Abidi, Direct-Conversion <strong>Radio</strong> Transceivers<strong>for</strong> Digital Communications, IEEE<br />
journal <strong>of</strong> solid-state circuits, Vol 30, no 12, December 1995<br />
[2] Blyakman A.B, Ryndyk A.G, Sidorov S.B, Forward Scattering Radar Moving Object<br />
Coordinate <strong>System</strong>, Radar, Sonar and Navigation, IEE Proceedings, Volume 152, Issue 3, 3 June<br />
2005 Page(s): 107 - 115<br />
[3] Chang K, RF and Microwave Wireless <strong>System</strong>s, Wiley Inter-Science, 2000<br />
[4] Del Mistro M, A Study <strong>of</strong> Bistatic Radar and the Development <strong>of</strong> an Independant Bistatic<br />
Receiver, University <strong>of</strong> Cape Town, 1992<br />
[5] Eric Blossom, GNU<strong>Radio</strong>: Tools <strong>for</strong> Exploring the <strong>Radio</strong> Frequency Spectrum,<br />
http://www.linuxjournal.com/article/7319, 2008<br />
[6] Ettus Reasearch LLC, TX and RX Daughterboards <strong>for</strong> the USRP S<strong>of</strong>tware <strong>Radio</strong> <strong>System</strong>,<br />
[7] Fette B.A., RF and Wireless Technologies,<br />
[8] Griffiths H D, Garnett A. J., Bistatic radar using satellite-borne illuminators <strong>of</strong> opportunity,<br />
Radar 92. International Conference, Volume , Issue , 12-13 Oct 1992 Page(s):276 - 279<br />
[9] Griffiths H.D., Bistatic and Multistatic Radar, University College London<br />
[10] Griffiths H.D., Baker C.J., Passive Coherent Location Radar <strong>System</strong>s. Part 1: Per<strong>for</strong>mance<br />
Prediction, Radar, Sonar and Navigation, IEE Proceedings, Volume 152, Issue 3, 3 June 2005<br />
Page(s): 153 - 159<br />
[11] Griffiths H.D., Baker C.J, Measurement and Analysis <strong>of</strong> Ambiguity Functions <strong>of</strong> Passive<br />
Radar Transmissions, Radar Conference, 2005 IEEE International, Volume , Issue , 9-12 May<br />
2005 Page(s): 321 - 325<br />
[12] Griffiths H.D., Long B.A., Television bases Bistatic Radar, IEE. Proceedings, Vol. 133,<br />
Part F, No.7, December 1986 Page(s) 649-657<br />
[13] Guner A., Temple M.A., Claypoole R.L. , Direct path filtering <strong>of</strong> DAB wave<strong>for</strong>m from PCL<br />
receiver target channel, Electronics Letters, Volume 39, Issue 1, 9 Jan. 2003 Page(s): 118 - 119<br />
[14] Hanle E, Ing Dr, Gunputh H.R., Survey <strong>of</strong> Bistatic and Multistatic Radar, IEE Proceedings,<br />
133(PtF No7):587– 595,Dec 1986.<br />
[15] Howland P.E. Television Based Bistatic Radar, University <strong>of</strong> Birmingham,1997<br />
[16] Howland P.E., Maksimiuk D., Reitsma G., FM based Bistatic Radar, Radar, Sonar and<br />
Navigation, IEE Proceedings, Volume 152, Issue 3, 3 June 2005 Page(s): 107 - 115<br />
[17] Jackson M.C, The Geometry <strong>of</strong> bistatic radar systems, Communications, Radar and Signal<br />
Processing, IEE Proceedings, Volume 133, Issue 7, December 1986 Page(s): 604 - 612<br />
[18] Jiabing Z., Liang T., Yi H., Study on Moving Target Detection to Passive Radar based on<br />
FM broadcast transmitter, Journal <strong>of</strong> <strong>System</strong>s Engineering and Electronics, Volume 18, Issue 3,<br />
2007, Pages 462-468<br />
[19] Kamas G., Lombardi M.A., Time and Frequecy Users Manual, NIST Special publication<br />
559, U.S. Government Printing Office, Washington DC,1990<br />
[20] Mitola J., S<strong>of</strong>tware <strong>Radio</strong> Architecture Evolution: Foundations, Technology Trade<strong>of</strong>fs, and<br />
Architecture Implications, IEICE Trans. Commun.E83-B1165–73<br />
[21] Mr Creese, Sentech Transmitter specification,<br />
[22] Nathanson F.E., Radar Design Principles, McGraw-Hill, 1969<br />
[23] Pozar D.M., Microwave and RF Design <strong>of</strong> Wireless <strong>System</strong>s, Wiley, 2001<br />
[24] Quegan S, Kingsley S, Understanding Radar <strong>System</strong>s, McGraw-Hill, 1992<br />
[25] Reed J. H., S<strong>of</strong>tware <strong>Radio</strong>: A Modern Approach to <strong>Radio</strong> Engineering, Prentice Hall, 2002<br />
[26] Sahr J.D., Lind F.D., The Manastash Ridge Radar: a passive bistatic radar <strong>for</strong> upper<br />
atmospheric radio science, <strong>Radio</strong> Science, Volume 32, 1997 Page(s) 2345 - 2358<br />
98
[27] Sandenbergh J.S, A GPS synchronised, Quart Crystal Frequency Standard, University <strong>of</strong><br />
Cape Town, 2005<br />
[28] schoenenberger, Principles <strong>of</strong> Independent Receivers <strong>for</strong> use with Cooperative Radar<br />
Transmitters, The <strong>Radio</strong> and Electronic Engineer, vol. 52, No. 2, February 1982 Page(s) 93-101<br />
[29] Skolnik M, Introduction to Radar <strong>System</strong>s, McGraw-Hill, 1962<br />
[30] Stein S.R., Frequency and Time, Their measurement and characterisation, Precision<br />
Frequency Control, Volume 2, E.A Gerber and A. Ballato, Eds., Academic Press, New<br />
York,1985 Page(s) 191 - 232<br />
[31] The S<strong>of</strong>tware <strong>Defined</strong> <strong>Radio</strong> Forum, http://www.sdr<strong>for</strong>um.org, 2008,<br />
[32] Watermeyer K, Design <strong>of</strong> a hardware plat<strong>for</strong>m <strong>for</strong> narrow-band S<strong>of</strong>tware <strong>Defined</strong> <strong>Radio</strong><br />
applications, University <strong>of</strong> Cape Town, 2007<br />
[33] Wikipedia The Free Encyclopedia, Passive Radar,<br />
http://en.wikipedia.org/wiki/Passive_radar, 2008<br />
[34] Willis N.J., Bistatic Radar, Artech House,1991<br />
[35] Willis N.J., Bistatic radars and their third resurgence: passive coherent location, IEEE<br />
Radar Conference, Long Beach, USA, 2002<br />
[36] Woodward P.M., Probability and In<strong>for</strong>mation Theory, with applications to Radar,<br />
Norwood, MA: Artech House, 1980<br />
[37] Microtune, 4937 D15 RF Tuner Module, 3x889 (3x7702), Advance Data Sheet, 2001<br />
[38] Analog Devices, AD9860/AD9862 Datasheet,<br />
[39] GNU<strong>Radio</strong> website, http://www.gnu.org/s<strong>of</strong>tware/gnuradio, 2008<br />
[40] Chang C.W.W, <strong>System</strong> Level Investigations <strong>of</strong> Television Based Bistatic Radar, University<br />
<strong>of</strong> Cape Town, 2005<br />
99
Appendix A<br />
Ambiguity Analysis Results<br />
Figure A.1: Ambiguity plot <strong>of</strong> <strong>Radio</strong> 2000 transmission<br />
100
Figure A.1 a: Zero doppler cut through ambiguity plot<br />
Figure A.1 b: Zero delay cut through ambiguity plot<br />
101
Table A.1: Bandwidth variation <strong>of</strong> <strong>Radio</strong> 2000 wave<strong>for</strong>m<br />
Time (s)<br />
Bandwidth (kHz)<br />
0.1 20<br />
0.2 6<br />
0.3 25<br />
0.4 10<br />
0.5 18.5<br />
0.6 31<br />
0.7 14<br />
0.8 5<br />
0.9 19<br />
1.0 15<br />
1.1 33<br />
1.2 0.5<br />
1.3 0.5<br />
1.4 18<br />
1.5 35<br />
1.6 23<br />
1.7 50<br />
1.8 5<br />
1.9 8<br />
2.0 16<br />
Average Bandwidth (kHz) 18.97<br />
102
Figure A.2: Ambiguity plot <strong>of</strong> Classic FM transmission<br />
Figure A.2 a: Zero doppler cut through ambiguity plot<br />
103
Figure A.2 b: Zero delay cut through ambiguity plot<br />
Table A.2: Bandwidth variation <strong>of</strong> Classic FM wave<strong>for</strong>m<br />
Time (s)<br />
Bandwidth (kHz)<br />
0.1 9<br />
0.2 7<br />
0.3 20<br />
0.4 25<br />
0.5 25<br />
0.6 15<br />
0.7 24<br />
0.8 22<br />
0.9 9<br />
1.0 32.5<br />
1.1 45<br />
1.2 13<br />
1.3 24<br />
1.4 48<br />
104
1.5 41<br />
1.6 13<br />
1.7 4.5<br />
1.8 0.6<br />
1.9 5<br />
2.0 4.5<br />
Average Bandwidth (kHz) 18.81<br />
Figure A.3: Ambiguity plot <strong>of</strong> UmhloboFM transmission<br />
105
Figure A.3 a: Zero doppler cut through ambiguity plot<br />
Figure A.3 b: Zero delay cut through ambiguity plot<br />
106
Table A.3: Bandwidth variation <strong>of</strong> Umhlobo FM wave<strong>for</strong>m<br />
Time (s)<br />
Bandwidth (kHz)<br />
0.1 41<br />
0.2 70<br />
0.3 75<br />
0.4 35<br />
0.5 53<br />
0.6 52<br />
0.7 36<br />
0.8 40<br />
0.9 45<br />
1.0 39<br />
1.1 42.5<br />
1.2 65<br />
1.3 28<br />
1.4 51<br />
1.5 35<br />
1.6 44<br />
1.7 23<br />
1.8 67<br />
1.9 68<br />
2.0 55<br />
Average Bandwidth (kHz) 45<br />
107
Figure A.4: Ambiguity plot <strong>for</strong> RGHP transmission<br />
Figure A.4 a: Zero doppler cut through ambiguity plot<br />
108
Figure A.4 b: Zero delay cut through ambiguity plot<br />
Table A.4: Bandwidth variation <strong>of</strong> RGHP wave<strong>for</strong>m<br />
Time (s)<br />
Bandwidth (kHz)<br />
0.1 39<br />
0.2 115<br />
0.3 73<br />
0.4 52<br />
0.5 38<br />
0.6 57<br />
0.7 46<br />
0.8 60<br />
0.9 53<br />
1.0 55<br />
1.1 107<br />
1.2 48<br />
1.3 49<br />
1.4 42<br />
1.5 59<br />
109
1.6 30<br />
1.7 63<br />
1.8 62<br />
1.9 50<br />
2.0 118<br />
Average Bandwidth (kHz) 63.6<br />
Figure A.5: Ambiguity plot <strong>of</strong> RSGR transmission<br />
110
Figure A.5 a: Zero doppler cut through ambiguity plot<br />
Figure A.5 b: Zero delay cut through ambiguity plot<br />
111
Table A.5: Bandwidth variation <strong>of</strong> RSGR wave<strong>for</strong>m<br />
Time (s)<br />
Bandwidth (kHz)<br />
0.1 3.5<br />
0.2 0.3<br />
0.3 4<br />
0.4 30<br />
0.5 9<br />
0.6 35<br />
0.7 40<br />
0.8 21<br />
0.9 18<br />
1.0 1.5<br />
1.1 55<br />
1.2 20<br />
1.3 50<br />
1.4 1.2<br />
1.5 4.5<br />
1.6 7.5<br />
1.7 5.5<br />
1.8 14<br />
1.9 5.5<br />
2.0 13<br />
Average Bandwidth (kHz) 16.1<br />
112
Figure A.6: Ambiguity plot <strong>of</strong> SAFM transmission<br />
113
Figure A.6 a: Zero doppler cut through ambiguity plot<br />
Figure A.6 b: Zero delay cut through ambiguity plot<br />
114
Table A.6: Bandwidth variation <strong>of</strong> SAFM wave<strong>for</strong>m<br />
Time (s)<br />
Bandwidth (kHz)<br />
0.1 40<br />
0.2 40<br />
0.3 47.5<br />
0.4 30<br />
0.5 47<br />
0.6 13<br />
0.7 17<br />
0.8 38<br />
0.9 2.5<br />
1.0 1.4<br />
1.1 2.5<br />
1.2 6<br />
1.3 18<br />
1.4 40<br />
1.5 6<br />
1.6 7.5<br />
1.7 1.4<br />
1.8 3.5<br />
1.9 1.6<br />
2.0 27.5<br />
Average Bandwidth (kHz) 19.9<br />
115
Appendix B<br />
Data Sheets and Sentech Table<br />
116
STATION NAME<br />
STATION<br />
CODE<br />
LAT<br />
DEG<br />
LAT<br />
MIN<br />
LAT<br />
SEC<br />
LONG<br />
DEG<br />
LONG<br />
MIN<br />
LONG<br />
SEC<br />
Sentech Station In<strong>for</strong>mation<br />
(First Page Only)<br />
MAP<br />
NUMBER<br />
SITE<br />
HEIGHT<br />
MAST<br />
HEIGHT<br />
SERVICE<br />
CODE<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 RAD5 89 1 10 FM 10 4 6<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 LOBO 92.1 1 10 FM 10 4 6<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 RGHP 95.3 1 10 FM 10 4 6<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 2000 98.6 1 10 FM 10 4 6<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 RSGR 102.1 1 10 FM 10 4 6<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 SAFM 105.7 1 10 FM 10 4 6<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 SBC1 183.25 1 15.85 TVV 12 4 3<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 SBC2 207.25 1 15.85 TVV 12 4 3<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 MNET 231.25 1 15.85 TVV 12 4 3<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 CSN 735.25 0.1 0.25 TVU 4 4 1<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 ETV 767.25 1 6.76 TVU 8.3 4 4<br />
CAPE TOWN C1 -34 -3 -15 18 23 15 3418AB 902 154 SBC3 799.25 1 6.76 TVU 8.3 4 4<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 METR 88.6 0.02 0.02 FM 0 1 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 RAD5 89.9 0.02 0.02 FM 0 1 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 LOBO 92.5 0.02 0.02 FM 0 1 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 RGHP 95.8 0.02 0.02 FM 0 1 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 2000 99.1 0.02 0.02 FM 0 1 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 RSGR 102.6 0.02 0.02 FM 0 1 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 SAFM 106.2 0.02 0.02 FM 0 1 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 SBC2 495.25 0.1 0.5 TVU 7 2 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 SBC1 527.25 0.1 0.5 TVU 7 2 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 MNET 591.25 0.1 0.5 TVU 7 2 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 SBC3 751.25 0.1 0.59 TVU 7.7 2 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 CSN 783.25 0.05 0.29 TVU 7.7 2 1<br />
TABLE MOUNTAIN C1.1 -33 -57 -25 18 24 13 3318CD 1067 14 ETV 815.25 0.1 0.59 TVU 7.7 2 1<br />
AURORA C1.10 -33 -49 -39 18 38 29 3318DC 176 23 SBC2 487.25 0 0 TVU 4.2 3 1<br />
AURORA C1.10 -33 -49 -39 18 38 29 3318DC 176 23 SBC3 551.25 0 0 TVU 4.2 3 1<br />
AURORA C1.10 -33 -49 -39 18 38 29 3318DC 176 23 MNET 583.25 0 0 TVU 4.2 3 1<br />
AURORA C1.10 -33 -49 -39 18 38 29 3318DC 176 23 SBC1 727.25 0 0 TVU 4.2 3 1<br />
AURORA C1.10 -33 -49 -39 18 38 29 3318DC 176 23 ETV 759.25 0 0 TVU 4.2 3 1<br />
GRABOUW C1.11 -34 -6 -5 18 58 3 3418BB 1181 30 RHLD 93.6 0.1 0.1 FM 0 1 1<br />
GRABOUW C1.11 -34 -6 -5 18 58 3 3418BB 1181 30 RKFM 94.9 0.01 0.01 FM 0 1 1<br />
GRABOUW C1.11 -34 -6 -5 18 58 3 3418BB 1181 30 RSGR 101.7 0.01 0.01 FM 0 1 1<br />
GRABOUW C1.11 -34 -6 -5 18 58 3 3418BB 1181 30 SAFM 105.3 0.01 0.01 FM 0 1 1<br />
GRABOUW C1.11 -34 -6 -5 18 58 3 3418BB 1181 30 HART 107.8 0 0.01 FM 4 1 1<br />
GRABOUW C1.11 -34 -6 -5 18 58 3 3418BB 1181 30 SBC2 615.25 0.05 0.5 TVU 10 3 4<br />
TX<br />
FREQ<br />
TX OUT<br />
PWR<br />
ERP<br />
SPEC<br />
FREQ<br />
BAND<br />
ANT<br />
GAIN<br />
ANT<br />
FACE<br />
ANT<br />
TIER
M I C R O T U N E<br />
4937 DI5 RF TUNER MODULE<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
CABLE MODEM APPLICATIONS<br />
1 APPLICATIONS<br />
The 4937 DI5 Tuner Module is specifically designed <strong>for</strong> subscriber-side cable<br />
modem applications.<br />
2 FEATURES<br />
Preliminary Preliminary and Confidential<br />
3 INTRODUCTION<br />
• DOCSIS compatible<br />
• VHF, Hyperband, and UHF<br />
• Band selection and tuning controlled by I 2 C bus<br />
• Downstream frequency range from 50 MHz to 860 MHz<br />
• Upstream frequency range from 5 MHz to 42 MHz<br />
• Single 5V power supply<br />
The receiver uses a single-conversion approach to 43.75 MHz with the reception<br />
frequency range divided into VHF low, VHF high, and UHF. A second conversion to<br />
5.75 MHz is available <strong>for</strong> QAM demodulators requiring a lower center frequency<br />
(3x7702); alternately, the output frequency is 43.75 MHz (3x8899).<br />
Figure 1<br />
4937 DI5 RF Tuner Modules<br />
Band selection and tuning is done via the I²C-bus, while a separate three-wire bus<br />
and transmit enable control the upstream amplifier.<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 1 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
The common cable input/output is realized by an F-connector (75Ω) per [IPS-sp-<br />
406].<br />
Two automatic gain control (AGC) inputs are available to level the signal into an<br />
external demodulator. The tuner’s intermediate frequency (IF) output is designed to<br />
drive a low-pass image reject filter prior to the QAM demodulator IC.<br />
A DC/DC converter is built in, so that only a single supply voltage <strong>of</strong> 5V is required.<br />
4 MECHANICAL SPECIFICATIONS<br />
This section contains mechanical specifications <strong>for</strong> the 4937 DI5 RF Tuner Module.<br />
4.1 MECHANICAL DRAWING<br />
Preliminary<br />
Figure 2<br />
Mechanical Drawing<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 2 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
4.2 MECHANICAL CHARACTERISTI CS<br />
Table 1<br />
Mechanical Characteristics<br />
CHARACTERISTIC<br />
DIMENSIONS<br />
Dimensions According to the drawing in Figure 2<br />
Weight<br />
Plug holding strength<br />
Tuner connection<br />
Screw fixing <strong>of</strong> F-connector ∗<br />
Approximately 56g<br />
Plug according to SCTE spec. IPS-sp-407<br />
The tuner provides four pins at bottom cover <strong>for</strong> horizontal<br />
mounting and grounding<br />
Absolute maximum torque strength: 3.39 Nm / only once<br />
Absolute maximum cantilever strength: 3.39 Nm<br />
Absolute maximum axial strength: 8.99N<br />
∗<br />
If the tuner is not mounted on the chassis, the frame may be bent during the test.<br />
Regardless <strong>of</strong> mounting, the F-connector will not be pulled out <strong>of</strong> the frame.<br />
5 FUNCTIONAL SPECIFICATIONS<br />
Preliminary<br />
5.1 ABSOLUTE MAXIMUM RATING S<br />
Stresses greater than those listed in Table 2 may cause permanent damage to the<br />
device. These are stress ratings only; functional operation <strong>of</strong> the device under<br />
conditions other than those listed in Table 3 is not recommended or implied.<br />
Exposure to any <strong>of</strong> the absolute-maximum rating conditions <strong>for</strong> extended periods <strong>of</strong><br />
time may affect reliability.<br />
Table 2<br />
Absolute Maximum Specifications<br />
PARAMETER MIN MAX UNIT<br />
Supply voltage 6 V<br />
AGC voltage 6 V<br />
Storage temperature -30 +70 °C<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 3 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
5.2 OPERATING CHARACTERISTICS<br />
The operating characteristics listed in Table 3 reflect the conditions necessary <strong>for</strong><br />
optimal per<strong>for</strong>mance and operating reliability.<br />
Table 3<br />
Operating Characteristics<br />
PARAMETER MIN TYP MAX UNIT<br />
CONDITIONS<br />
OR LOCATION<br />
Frequency range<br />
VHF Low 50 162 MHz<br />
VHF High 156 469 MHz<br />
UHF 463 860 MHz<br />
Frequency range, referenced to center<br />
frequency <strong>of</strong> 6 MHz bandwidth<br />
VHF Low 53 159 MHz<br />
VHF High 159 466 MHz<br />
UHF 466 857 MHz<br />
Tuning resolution<br />
Preliminary<br />
Standard tuning increment (see<br />
Table 8)<br />
62.5 kHz<br />
Recommended takeover frequencies,<br />
referred to center frequency<br />
VHF Low / VHF High 158 MHz<br />
UHF 464 MHz<br />
Output Frequency<br />
3x7702 5.75 MHz ± 0.05 MHz<br />
3x8899 43.75 MHz ± 0.05 MHz<br />
Input impedance<br />
VHF/UHF Common 75 Ω Unbalanced<br />
AGC voltage <strong>for</strong> maximum gain<br />
RF 4 V ± 0.1V<br />
IF 4 V ± 0.1V<br />
Power supply voltage<br />
Voltage V S1 5 ± 0.3 V Pin 3<br />
Voltage condition V S1 150 mA<br />
Voltage V S2 5 ±0.25 V Pin 6<br />
Voltage condition V S2 200 mA<br />
Voltage V S3 5 ±0.25 V Pin 10<br />
Voltage condition V S3 200 mA<br />
Voltage V S4 5 ±0.3 V Pin 15<br />
Voltage condition V S4 100 mA<br />
Permissible ripple voltage (20 Hz to 100<br />
kHz)<br />
20 mVpp<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 4 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
PARAMETER MIN TYP MAX UNIT<br />
CONDITIONS<br />
OR LOCATION<br />
Temperature<br />
Operating temperature 0 60 °C<br />
6 TUNER DOWNSTREAM DATA<br />
Table 4<br />
Electrical Characteristics<br />
PARAMETER TEST CONDITIONS MIN TYP MAX UNIT<br />
Frequency range 55 860 MHz<br />
Preliminary<br />
Input signal level 40 80 dBµV<br />
Voltage gain<br />
Output level at 1 kΩ<br />
Measured between antenna input and<br />
IF output (pins 17 and 18). The input is<br />
loaded with 75Ω and the IF output is<br />
loaded with a test circuit (see Figure 5).<br />
The output impedance is about 220Ω.<br />
Pins 17 and 18 are not DC decoupled.<br />
60 80 95 dB<br />
1 Vpp<br />
VHF Low 8 10 dB<br />
Noise figure<br />
VHF High 8 10 dB<br />
UHF 8 10 dB<br />
VSWR Antenna input 3<br />
IF Rejection<br />
[Rejection <strong>of</strong> CW<br />
Signal at highest<br />
VHF Low 50 70 dB<br />
possible IF (46.75<br />
MHz) fed into the tuner<br />
input relative to a CW<br />
at desired channel<br />
center frequency<br />
measured at the IF<br />
mixer output. Both<br />
signals must have the<br />
same level at F-<br />
connector input.]<br />
VHF High<br />
UHF<br />
60<br />
60<br />
80<br />
80<br />
dB<br />
dB<br />
Upstream rejection<br />
Image rejection<br />
RF Tilt<br />
Signal level <strong>for</strong> 1 dB<br />
gain compression<br />
Phase noise<br />
Isolation between upstream output<br />
(5 MHz to 42 MHz) and IF mixer out<br />
(40.75 MHz to 46.75 MHz)<br />
75 dB<br />
VHF Low 60 70 dB<br />
VHF High 55 65 dB<br />
UHF 55 60 dB<br />
For all AGC settings and over a 6 MHz<br />
bandwidth around center frequency<br />
AGC deactivated with AGC = 4V (pins<br />
7 and 16) <strong>for</strong> maximum gain<br />
2.5 dB<br />
72 dBµV<br />
VHF Low -71 -55 dBc/Hz<br />
VHF High<br />
Measured at 1 kHz distance from<br />
carrier<br />
-60 -55 dBc/Hz<br />
UHF<br />
-58 -55 dBc/Hz<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 5 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
PARAMETER TEST CONDITIONS MIN TYP MAX UNIT<br />
VHF Low -95 -80 dBc/Hz<br />
VHF High<br />
Measured at 10 kHz distance from<br />
carrier<br />
-85 -80 dBc/Hz<br />
UHF<br />
-85 -80 dBc/Hz<br />
VHF Low -102 -90 dBc/Hz<br />
VHF High<br />
Measured at 20 kHz distance from<br />
carrier<br />
-92 -85 dBc/Hz<br />
UHF<br />
-90 -85 dBc/Hz<br />
VHF Low -109 -100 dBc/Hz<br />
VHF High<br />
Measured at 100 kHz distance from<br />
carrier<br />
-106 -100 dBc/Hz<br />
UHF<br />
-103 -100 dBc/Hz<br />
Oscillator voltage F-connector terminated with 75Ω<br />
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
6.1 INFLUENCE OF AGC<br />
The curves in Figure 3 and Figure 4 are measured at +25°C with an input level <strong>of</strong><br />
45 dBµV. The values are typical values and can vary within the guaranteed limits.<br />
60,0<br />
50,0<br />
40,0<br />
30,0<br />
RF gain [dB]<br />
20,0<br />
10,0<br />
VHF-low (100MHz)<br />
VHF-high (350MHz)<br />
UHF (650MHz)<br />
0,0<br />
0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5<br />
-10,0<br />
-20,0<br />
RF-AGC voltage [V]<br />
Preliminary<br />
35,0<br />
30,0<br />
25,0<br />
Figure 3<br />
RF Gain vs. AGC Voltage<br />
20,0<br />
IF gain [dB]<br />
15,0<br />
10,0<br />
all bands<br />
5,0<br />
0,0<br />
0,0 0,5 1,0 1,5 2,0 2,5 3,0 3,5 4,0 4,5<br />
-5,0<br />
IF-AGC voltage [V]<br />
Figure 4<br />
IF Gain vs. AGC Voltage<br />
The noise figure shall not increase by more than the corresponding AGC gain<br />
reduction. The input return loss shall be maintained within the specified limits over<br />
the entire range <strong>of</strong> AGC voltage.<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 7 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
7 TUNER UPSTREAM DATA<br />
All data is measured according to the test circuit shown in Figure 6 on page 9. The<br />
input impedance between Pins 1 and 2 <strong>for</strong> this tuner is 50 ohms.<br />
Table 5<br />
Tuner Upstream Data<br />
PARAMETER TEST CONDITIONS MIN TYP MAX UNIT<br />
Input level Source impedance 75Ω sym 33 35 dBmV<br />
Preliminary<br />
Voltage gain<br />
Gain control word = maximum<br />
gain<br />
25 27 29 dB<br />
Gain steps 0.7 1 1.3 dB<br />
Gain range 59 dB<br />
Group delay variation<br />
Amplitude ripple variation<br />
5 MHz to 42 MHz (3.2 MHz<br />
bandwidth)<br />
60 nsec<br />
5 MHz to 42 MHz 1.28 MHz bandwidth ± 0.2 dB<br />
Absolute accuracy <strong>of</strong><br />
transmitted power<br />
TX Transient Spurs<br />
Gain setting = maximum<br />
gain<br />
Gain setting < (maximum<br />
gain –12)<br />
5 MHz to 42 MHz ± 2 dB<br />
16 mVp-p<br />
8 mVp-p<br />
TX Transient duration TXEN rise/fall time < 0.1 µs 2 µsec<br />
Reverse channel harmonic<br />
distortion<br />
V out = +58 dBmV<br />
5 MHz to 42 MHz 2 nd harmonic level, single tone -53 dBc<br />
5 MHz to 42 MHz 3 rd harmonic level, single tone -54 dBc<br />
54 MHz to 60 MHz -40 -35 dBmV<br />
60 MHz to 88 MHz -50 -40 dBmV<br />
88 MHz to 860 MHz -50 -45 dBmV<br />
Noise floor<br />
Input terminated with 75Ω<br />
Transmit mode noise Voltage gain 24 dB 131 150 nV / √Hz<br />
Transmit disable mode<br />
noise<br />
TXEN low, voltage gain 24 dB 810 pV / √Hz<br />
8 TUNER MEASUREMENT TEST CO NDITIONS<br />
All tuner data are held under the following conditions unless otherwise noted:<br />
• Measurement tolerance 10% or 1 dB<br />
• Ambient temperature + 25°C ± 3°C<br />
• Supply voltages + 5V ± 2%<br />
• AGC voltage + 4V ± 2%<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 8 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
8.1 TEST CIRCUITS<br />
8.1.1 VOLTAGE GAIN, TILT, AND NO ISE FIGURE<br />
Tuner<br />
Pin 17<br />
Pin 18<br />
680 Ω<br />
680 =Ω<br />
4:1<br />
T1<br />
75 Ω=Impedance<br />
Figure 5<br />
Test Circuit <strong>for</strong> Voltage Gain, Tilt, and Noise Figure<br />
Preliminary<br />
For the voltage gain, tilt, and noise figure test circuit:<br />
• Loss <strong>of</strong> test-dummy: 22.6 dB<br />
• T1 = RF – Trans<strong>for</strong>mer (ohms - ratio = 1:4)<br />
• Type: MCL T4-1 or equivalent<br />
8.1.2 UPSTREAM CHANNEL<br />
Tuner<br />
Pin 1<br />
Pin 2<br />
1:1<br />
T1<br />
75 Ω=Impedance<br />
Figure 6<br />
Test Circuit <strong>for</strong> Upstream Channel<br />
For the upstream channel test circuit:<br />
• Loss <strong>of</strong> test-dummy: < 1 dB<br />
• T1 = RF – Trans<strong>for</strong>mer (ohms - ratio = 1:1)<br />
• Type: MCL T1-1 or equivalent<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 9 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
9 CONTROL<br />
9.1 WRITE DATA FORMAT FOR I 2 C BUS<br />
Table 6<br />
Write Data Format<br />
MSB LSB ACK<br />
Address byte 1 1 0 0 0 MA1 MA0 R/W 1 A 2<br />
Divider byte 1 0 N14 N13 N12 N11 N10 N9 N8 A<br />
Divider byte 2 N7 N6 N5 N4 N3 N2 N1 N0 A<br />
Control byte 1 1 CP T2 T1 T0 RSA RSB OS A<br />
Control byte 2 P7 P6 P5 P4 P3 P2 P1 P0 A<br />
1<br />
R/W = 0 is write mode<br />
2<br />
A = Acknowledge<br />
9.2 ADDRESS SELECTION FOR I 2 C B US<br />
Preliminary<br />
Table 7 Address Selection<br />
MA1 MA0 ADDRESS VOLTAGE AT PIN 11<br />
0 0 C0 (0 to 0.1) V S3<br />
0 1 C2 Open circuit or (0.2 to 0.3) V S3<br />
1 0 C4 (0.4 to 0.6) V S3<br />
1 1 C6 (0.9 to 1) V S3<br />
9.3 OSCILLATOR FREQUENCY AND DIVIDER BYTE CALCULATION<br />
Table 8<br />
Oscillator Frequency and Divider Byte Calculation<br />
RSA RSB REFERENCE DIVIDER<br />
MINIMUM TUNING<br />
STEP<br />
F REF<br />
1 1 512 62.5 kHz 7.8125 kHz<br />
X 0 640 50.0 kHz 6.25 kHz<br />
0 1 1024 31.25 kHz 3.90625 kHz<br />
Use the following <strong>for</strong>mula to calculate oscillator frequency and divider byte.<br />
f osc = f ref × 8 × SF<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 10 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
Where:<br />
f osc = Local oscillator frequency<br />
f ref = Crystal reference frequency / 512 = 4 MHz / 512 = 7.8125 kHz<br />
SF = Programmable scaling factor<br />
Scaling factor is SF = 16384 × n14 + 8192 × n13 + 4096 × n12 + 2048 ×<br />
n11 + 1024 × n10 + 512 × n9 + 256 × n8 + 128 × n7 + 64 × n6 + 32 × n5<br />
+ 16 × n4 + 8 × n3 + 4 × n2 + 2 × n1 + n0<br />
9.4 CONTROL BYTE (I 2 C)<br />
Table 9<br />
Control Byte 1 Settings (Default)<br />
MSB LSB ACK<br />
Control byte 1 1 0 0 0 1 1 1 0 A<br />
Table 10<br />
Control Byte 1 Settings Default Descriptions<br />
CODE DESCRIPTION SETTINGS<br />
Preliminary<br />
CP<br />
Charge pump current<br />
1 = Fastest tuning<br />
0 = Better phase noise <strong>for</strong> distance < 10<br />
kHz to the carrier<br />
OS Tuning voltage<br />
0 = On<br />
1 = Off<br />
RSA, RSB Reference divider See Table 8 on page 10<br />
T0, T1, T2 Test mode bit See Table 11<br />
Table 11<br />
Test Mode Bit Settings<br />
T2 T1 T0 DEVICE OPERATION<br />
0 0 1 Normal mode<br />
0 1 x Charge pump is <strong>of</strong>f<br />
1 1 0 Charge pump is sinking current<br />
1 1 1 Charge pump is sourcing current<br />
1 0 0 Internal test mode<br />
1 0 1 Internal test mode<br />
Table 12<br />
Control Byte 2 (Band Selection)<br />
BAND ACTIVE PORT P7 P6 P5 P4 P3 P2 P1 P0<br />
UHF P0 0 X 1 1 X X X X<br />
VHF High P2 1 X 0 1 X X X X<br />
VHF Low P1 1 X 1 0 X X X X<br />
Note: X = not used, P3 = used <strong>for</strong> upstream shutdown (see section 9.6)<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 11 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
9.5 READ DATA FORMAT (I2C)<br />
Table 13 Read Data Format (I 2 C)<br />
MSB LSB ACK<br />
Address byte 1 1 0 0 0 MA1 MA0 R/W A<br />
Status byte POR FL I2 I1 I0 A2 A1 A0 A<br />
Note: MSB is transmitted first.<br />
Table 14<br />
Read Data Format Descriptions<br />
CODE<br />
DESCRIPTION<br />
R/W 1 = Read mode<br />
POR Power on reset flag (POR = 1 at power on)<br />
FL<br />
In lock flag (FL = 1 when PLL is locked)<br />
I2, I1, I0 Digital levels <strong>for</strong> I/O ports P0, P1, and P2<br />
A2, A1, A0<br />
Digital output <strong>of</strong> 5-level ADC <strong>for</strong> AFC function. Values<br />
<strong>for</strong> correct tuning: A2 = 0, A1= 1, A0 = 0<br />
Preliminary<br />
9.6 PROGRAMMABLE-GAIN AMPLI FIER CONTROL (THREE-WIRE BUS)<br />
Table 15 Pin Map (Three-Wire Bus)<br />
PIN SYMBOL DESCRIPTION<br />
4 AS1 Active low enable<br />
5 TXEnable Hardware shutdown<br />
8 SCL1 Serial clock<br />
9 SDA1 Serial data<br />
A serial data interface controls the programmable-gain amplifier (PGA). It has an<br />
active-low enable (AS1) to sample the data, with data clocked in MSB (D7) first on<br />
the rising edge <strong>of</strong> SCL1. Data is stored on the rising edge <strong>of</strong> AS1. The gain is<br />
determined by a 6-bit word (D5 – D0).<br />
Table 16<br />
Data Register (3-Wire Bus)<br />
BIT MNEMONIC DESCRIPTION<br />
MSB 7 D7 S<strong>of</strong>tware shutdown<br />
6 D6 Test bit<br />
5 D5 Gain control, bit 5<br />
4 D4 Gain control, bit 4<br />
3 D3 Gain control, bit 3<br />
2 D2 Gain control, bit 2<br />
1 D1 Gain control, bit 1<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 12 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
BIT MNEMONIC DESCRIPTION<br />
0 D0 Gain control, bit 0<br />
Setting PLL-Port 3 low shuts down the PGA. Port 3 is controlled over the I 2 C bus<br />
(SDA2; SCL2). Control byte 2 (P3) has to be 1 <strong>for</strong> shutdown or 0 <strong>for</strong> normal mode.<br />
Hardware shutdown overrides s<strong>of</strong>tware shutdown (D7) and stored gain settings will<br />
be lost. In normal active mode, port 3 must be held high. To bias only the differential<br />
output-power-amp between bursts, TXEnable (Pin 5) must be held low. TXEnable<br />
must be held high <strong>for</strong> transmit mode.<br />
Table 17<br />
State Diagram (3-Wire Bus)<br />
SHDN<br />
PORT 3<br />
TXEN<br />
PIN 5 D7 D6 D5 D4 D3 D2 D1 D0 STATE<br />
1 0 X X X X X X X X Shutdown mode<br />
0 0 0 X X X X X X X S<strong>of</strong>tware shutdown mode<br />
0 0 1 X X X X X X X Transmit disable mode<br />
0 1 1 X X X X X X X Transmit mode<br />
Preliminary<br />
0 1 1 X 0 0 0 0 0 0<br />
Maximum gain – 63 dB =<br />
minimum gain<br />
0 1 1 X 0 0 0 0 0 1 Maximum gain – 62 dB<br />
0 1 1 X - - - - - - -<br />
0 1 1 X 1 0 0 0 0 1 Maximum gain – 30 dB<br />
0 1 1 X - - - - - - -<br />
0 1 1 X 1 1 1 1 1 0 Maximum gain – 1 dB<br />
0 1 1 X 1 1 1 1 1 1 Maximum gain<br />
9.7 SERIAL INTERFACE TIMING<br />
A G B C D E F<br />
AS1<br />
SCL1<br />
D7 D6 D5 D4 D3 D2 D1 D0<br />
SDA1<br />
Figure 7<br />
Serial Interface Timing<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 13 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
Table 18<br />
Timing Characteristics<br />
PARAMETER SYMBOL MIN TYP MAX UNITS<br />
AS1 to SCL1 rise setup time A 10 ns<br />
AS1 to SCL1 rise hold time F 20 ns<br />
SDA1 to SCL1 setup time B 10 ns<br />
SDA1 to SCL1 hold time C 20 ns<br />
SDA1 pulse width high G 50 ns<br />
SDA1 pulse width low G 50 ns<br />
SCL1 pulse width high E 50 ns<br />
SCL1 pulse width low D 50 ns<br />
10 SAFETY AND RELIABILITY<br />
10.1 ELECTROSTATIC DISCHARGE (E SD) PROTECTION<br />
Preliminary<br />
10.2 HIGH VOLTAGE<br />
WARNING:<br />
Observe these precautions:<br />
The 4937 DI5 Tuner Module contains components that can be<br />
damaged by electrostatic discharge.<br />
• Ground yourself be<strong>for</strong>e handling the tuner.<br />
• Do not touch the tuner connector pins without ESD protection.<br />
The tuner meets specifications IEC 801.2 level 2.<br />
10.3 HUMIDITY<br />
Table 19<br />
Local Oscillator Drift<br />
PARAMETER DRIFT UNIT PROCEDURE<br />
VHF Low ± 15 kHz<br />
VHF High ± 45 kHz<br />
UHF ± 75 kHz<br />
1. Run 60 hours at 55°C and 20% relative<br />
humidity.<br />
2. Run 1 hour at 23°C and 50% relative<br />
humidity.<br />
3. Take first measurement.<br />
4. Run 65 hours at +40°C and 95% relative<br />
humidity.<br />
5. Take second measurement.<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 14 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
10.4 VIBRATION TEST<br />
After applying vibration <strong>of</strong> 1.5 mm amplitude, frequency <strong>of</strong> 10 - 55 -10 Hz (1 minute)<br />
each X, Y, Z direction <strong>for</strong> 2 hours (total 6 hours), tuner shall not have any rattling or<br />
loosening and shall comply with the variation to its initial value as listed in Table 20.<br />
Table 20<br />
Vibration Test<br />
PARAMETER MEASUREMENT UNIT<br />
Gain variation < ± 3 dB<br />
Wave variation < ± 30 %<br />
10.5 MICROPHONY<br />
The microphony test is made with a TV set. The resolution is optimal. With maximum<br />
AF output <strong>of</strong> the TV set, the tuner is free <strong>of</strong> microphonic effects, provided the unit is<br />
installed in a pr<strong>of</strong>essional manner.<br />
Preliminary<br />
10.6 LOOSE CONTACT TEST OF TUNE R ALONE<br />
10.7 SOLDER LIMITS<br />
The test pattern is a color bar. The resolution is 3 MHz. To test, there must be no<br />
interruption effects when the edge <strong>of</strong> the tuner is knocked, provided it is fastened<br />
with a ground contact.<br />
See application note APN001.<br />
10.8 NATIONAL REGULATIONS<br />
The tuner meets the requirements <strong>of</strong> VDE 9872/7.72 and Amtsblatt DBP 069/1981<br />
(FTZ), EN 55013, EN 55020 (if properly mounted into TV set, VCR, or converter).<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 15 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
11 ORDERING INFORMATION<br />
The 4937 DI5 Tuner Modules may be ordered in the packaging units and quantities<br />
shown in Table 21 and Table 22. For packaging options and quantities other than<br />
those shown, contact one <strong>of</strong> the <strong>of</strong>fices listed on the last page <strong>of</strong> this document.<br />
Table 21<br />
Packaging Units<br />
4937 TUNER MODELS<br />
PACKAGING UNITS 3X8899 3X7702<br />
Number <strong>of</strong> Tuner Modules Per Box 72 72<br />
Number <strong>of</strong> Boxes Per Master Box 40 40<br />
Table 22<br />
Order Quantities<br />
TOTAL NUMBER OF TUNERS PER<br />
MASTER BOX<br />
NUMBER OF MASTER<br />
BOXES 3X8899 3X7702<br />
Preliminary<br />
0.5 1,440 1,440<br />
1.0 2,880 2,880<br />
1.5 4,320 4,320<br />
2.0 5,760 5,760<br />
2.5 7,200 7,200<br />
3.0 8,640 8,640<br />
3.5 10,080 10,080<br />
4.0 11,520 11,520<br />
4.5 12,960 12,960<br />
5.0 14,400 14,400<br />
12 REVISION HISTORY<br />
NAME<br />
DESCRIPTION<br />
ECN<br />
NO. DATE REV<br />
Hennig 24.11.00 M1<br />
Hennig 011/01 20.02.01 01<br />
Hennig Change 3x7702 (3x8899) to 3x8899 (3x7702) 050/01 10.07.01 02<br />
3X 8899 Revision: 02<br />
September 01<br />
Page 16 <strong>of</strong> 17<br />
3x8899_02.doc
4937 DI5 TUNER MODULE CABLE MODEM APPLICATIONS<br />
3X8899 (3X7702)<br />
ADVANCE DATA SHEET<br />
NOTICES<br />
Preliminary<br />
Preliminary<br />
and Confidential<br />
NOTICE - The in<strong>for</strong>mation in this document is believed to be accurate and reliable. Microtune assumes no<br />
responsibility <strong>for</strong> any consequences arising from the use <strong>of</strong> this in<strong>for</strong>mation, nor from any infringement <strong>of</strong> patents<br />
or the rights <strong>of</strong> third parties which may result from its use. No license is granted by implication or otherwise under<br />
any patent or other rights <strong>of</strong> Microtune. The in<strong>for</strong>mation in this publication replaces and supersedes all<br />
in<strong>for</strong>mation previously supplied, and is subject to change without notice. The customer is responsible <strong>for</strong><br />
assuring that proper design and operating safeguards are observed to minimize inherent and procedural<br />
hazards. Microtune assumes no responsibility <strong>for</strong> applications assistance or customer product design.<br />
NOTICE - The devices described in this document are not authorized <strong>for</strong> use in medical, life-support equipment,<br />
or any other application involving a potential risk <strong>of</strong> severe property or environmental damage, personal injury, or<br />
death without prior express written approval <strong>of</strong> Microtune. Any such use is understood to be entirely at the user’s<br />
risk.<br />
TRADEMARKS - Microtune, MicroTuner, and the Microtune logo are trademarks <strong>of</strong> Microtune, Inc. All other<br />
trademarks belong to their respective companies.<br />
PATENTS – Microtune’s products are protected by one or more <strong>of</strong> the following U.S. patents: 5,625,325;<br />
5,648,744; 5,717,730; 5,737,035; 5,739,730; 5,805,988; 5,847,612; 6,100,761; 6,104,242; 6,144,402; 6,163,684;<br />
6,169,569; 6,177,964; 6,218,899 and additional patents pending or filed.<br />
COPYRIGHT - Entire contents Copyright © 2001 Microtune, Inc.<br />
World Headquarters<br />
Microtune, Inc.<br />
2201 Tenth Street<br />
Plano, TX 75074<br />
USA<br />
Telephone: 972-673-1600<br />
Fax: 972-673-1602<br />
Email: sales@microtune.com<br />
Website: www.microtune.com<br />
European Headquarters<br />
Microtune GmbH and Co. KG<br />
Marie Curie Strasse 1<br />
85055 Ingolstadt / Germany<br />
Telephone: +49-841-9378-011<br />
Fax: +49-841-9378-010<br />
Sales Telephone: +49-841-9378-020<br />
Sales Fax: +49-841-9378-024<br />
Pan-Asian Headquarters<br />
Microtune, Inc. - Hong Kong<br />
Silvercord Tower 1, Room 503<br />
30 Canton Road<br />
Kowloon, Hong Kong<br />
Telephone: +852-2378-8128<br />
Fax: +852-2302-0756<br />
For a detailed list <strong>of</strong> current sales representatives, visit our Web site at www.microtune.com.<br />
3X 8899 Revision: 02<br />
August 01<br />
Page 17 <strong>of</strong> 17<br />
3x8899_02.doc